harivansh-afk/sandbox-agent
1
0
Fork 0
mirror of https://github.com/harivansh-afk/sandbox-agent.git synced 2026-04-15 06:04:43 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat: acp http adapter

Browse source
This commit is contained in:
Nathan Flurry 2026-02-10 16:05:56 -08:00
parent 2ba630c180
commit b4c8564cb2
217 changed files with 18785 additions and 17400 deletions
Download patch file Download diff file Expand all files Collapse all files

2
.claude/commands/post-release-testing.md
View file

@ -43,7 +43,7 @@ Manually verify the install script works in a fresh environment:
```bash
docker run --rm alpine:latest sh -c "
apk add --no-cache curl ca-certificates libstdc++ libgcc bash &&
curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh &&
curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh &&
sandbox-agent --version
"
```

8
.github/workflows/ci.yaml vendored
View file

@ -14,7 +14,7 @@ jobs:
- uses: dtolnay/rust-toolchain@stable
with:
components: rustfmt, clippy
- uses: Swatinem/rust-cache@v2
- uses: Swatinem/rust-cache@main
- uses: pnpm/action-setup@v4
- uses: actions/setup-node@v4
with:
@ -23,11 +23,11 @@ jobs:
- run: pnpm install
- name: Run checks
run: ./scripts/release/main.ts --version 0.0.0 --check
- name: Run ACP v2 server tests
- name: Run ACP v1 server tests
run: |
cargo test -p sandbox-agent-agent-management
cargo test -p sandbox-agent --test v2_api
cargo test -p sandbox-agent --test v2_agent_process_matrix
cargo test -p sandbox-agent --test v1_api
cargo test -p sandbox-agent --test v1_agent_process_matrix
cargo test -p sandbox-agent --lib
- name: Run SDK tests
run: pnpm --dir sdks/typescript test

4
.github/workflows/release.yaml vendored
View file

@ -147,8 +147,8 @@ jobs:
sudo apt-get install -y unzip curl
# Install AWS CLI
curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscliv2.zip"
unzip awscliv2.zip
curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscli.zip"
unzip awscli.zip
sudo ./aws/install --update
COMMIT_SHA_SHORT="${GITHUB_SHA::7}"

94
CLAUDE.md
View file

@ -1,24 +1,24 @@
# Instructions
## ACP v2 Baseline
## ACP v1 Baseline
- v2 is ACP-native.
- v1 is ACP-native.
- `/v1/*` is removed and returns `410 Gone` (`application/problem+json`).
- `/opencode/*` is disabled during ACP core phases and returns `503`.
- Prompt/session traffic is ACP JSON-RPC over streamable HTTP on `/v2/rpc`:
- `POST /v2/rpc`
- `GET /v2/rpc` (SSE)
- `DELETE /v2/rpc`
- Prompt/session traffic is ACP JSON-RPC over streamable HTTP on `/v1/rpc`:
- `POST /v1/rpc`
- `GET /v1/rpc` (SSE)
- `DELETE /v1/rpc`
- Control-plane endpoints:
- `GET /v2/health`
- `GET /v2/agents`
- `POST /v2/agents/{agent}/install`
- `GET /v1/health`
- `GET /v1/agents`
- `POST /v1/agents/{agent}/install`
- Binary filesystem transfer endpoints (intentionally HTTP, not ACP extension methods):
- `GET /v2/fs/file`
- `PUT /v2/fs/file`
- `POST /v2/fs/upload-batch`
- `GET /v1/fs/file`
- `PUT /v1/fs/file`
- `POST /v1/fs/upload-batch`
- Sandbox Agent ACP extension method naming:
- Custom ACP methods use `_sandboxagent/...` (not `_sandboxagent/v2/...`).
- Custom ACP methods use `_sandboxagent/...` (not `_sandboxagent/v1/...`).
- Session detach method is `_sandboxagent/session/detach`.
## API Scope
@ -27,7 +27,7 @@
- ACP extensions may be used for gaps (for example `skills`, `models`, and related metadata), but the default is that agent-facing behavior is implemented by the agent through ACP.
- Custom HTTP APIs are for non-agent/session platform services (for example filesystem, terminals, and other host/runtime capabilities).
- Filesystem and terminal APIs remain Sandbox Agent-specific HTTP contracts and are not ACP.
- Keep `GET /v2/fs/file`, `PUT /v2/fs/file`, and `POST /v2/fs/upload-batch` on HTTP:
- Keep `GET /v1/fs/file`, `PUT /v1/fs/file`, and `POST /v1/fs/upload-batch` on HTTP:
- These are Sandbox Agent host/runtime operations with cross-agent-consistent behavior.
- They may involve very large binary transfers that ACP JSON-RPC envelopes are not suited to stream.
- This is intentionally separate from ACP native `fs/read_text_file` and `fs/write_text_file`.
@ -51,14 +51,24 @@
## TypeScript SDK Architecture
- TypeScript clients are split into:
- `acp-http-client`: protocol-pure ACP-over-HTTP (`/v2/rpc`) with no Sandbox-specific metadata/extensions.
- `sandbox-agent`: `SandboxAgentClient` wrapper that adds Sandbox metadata/extension helpers and keeps non-ACP HTTP helpers.
- `SandboxAgentClient` constructor is `new SandboxAgentClient(...)`.
- `SandboxAgentClient` auto-connects by default; `autoConnect: false` requires explicit `.connect()`.
- ACP/session methods must throw when disconnected (`NotConnectedError`), and `.connect()` must throw when already connected (`AlreadyConnectedError`).
- A `SandboxAgentClient` instance may have at most one active ACP connection at a time.
- Stable ACP session method names should stay ACP-aligned in the Sandbox wrapper (`newSession`, `loadSession`, `prompt`, `cancel`, `setSessionMode`, `setSessionConfigOption`).
- Sandbox extension methods are first-class wrapper helpers (`listModels`, `setMetadata`, `detachSession`, `terminateSession`).
- `acp-http-client`: protocol-pure ACP-over-HTTP (`/v1/acp`) with no Sandbox-specific HTTP helpers.
- `sandbox-agent`: `SandboxAgent` SDK wrapper that combines ACP session operations with Sandbox control-plane and filesystem helpers.
- `SandboxAgent` entry points are `SandboxAgent.connect(...)` and `SandboxAgent.start(...)`.
- Stable Sandbox session methods are `createSession`, `resumeSession`, `resumeOrCreateSession`, `destroySession`, `sendSessionMethod`, and `onSessionEvent`.
- `Session` helpers are `prompt(...)`, `send(...)`, and `onEvent(...)`.
- Cleanup is `sdk.dispose()`.
### Docs Source Of Truth
- For TypeScript docs/examples, source of truth is implementation in:
- `sdks/typescript/src/client.ts`
- `sdks/typescript/src/index.ts`
- `sdks/acp-http-client/src/index.ts`
- Do not document TypeScript APIs unless they are exported and implemented in those files.
- For HTTP/CLI docs/examples, source of truth is:
- `server/packages/sandbox-agent/src/router.rs`
- `server/packages/sandbox-agent/src/cli.rs`
- Keep docs aligned to implemented endpoints/commands only (for example ACP under `/v1/acp`, not legacy `/v1/sessions` APIs).
## Source Documents
@ -76,5 +86,43 @@
- Regenerate `docs/openapi.json` when HTTP contracts change.
- Keep `docs/inspector.mdx` and `docs/sdks/typescript.mdx` aligned with implementation.
- Append blockers/decisions to `research/acp/friction.md` during ACP work.
- TypeScript SDK tests should run against a real running server/runtime over real `/v2` HTTP APIs, typically using the real `mock` agent for deterministic behavior.
- TypeScript SDK tests should run against a real running server/runtime over real `/v1` HTTP APIs, typically using the real `mock` agent for deterministic behavior.
- Do not use Vitest fetch/transport mocks to simulate server functionality in TypeScript SDK tests.
## Docker Examples (Dev Testing)
- When manually testing bleeding-edge (unreleased) versions of sandbox-agent in `examples/`, use `SANDBOX_AGENT_DEV=1` with the Docker-based examples.
- This triggers `examples/shared/Dockerfile.dev` which builds the server binary from local source and packages it into the Docker image.
- Example: `SANDBOX_AGENT_DEV=1 pnpm --filter @sandbox-agent/example-mcp start`
## Install Version References
- Channel policy:
- Sandbox Agent install/version references use a pinned minor channel `0.N.x` (for curl URLs and `sandbox-agent` / `@sandbox-agent/cli` npm/bun installs).
- Gigacode install/version references use `latest` (for `@sandbox-agent/gigacode` install/run commands and `gigacode-install.*` release promotion).
- Release promotion policy: `latest` releases must still update `latest`; when a release is `latest`, Sandbox Agent must also be promoted to the matching minor channel `0.N.x`.
- Keep every install-version reference below in sync whenever versions/channels change:
- `README.md`
- `docs/acp-http-client.mdx`
- `docs/cli.mdx`
- `docs/quickstart.mdx`
- `docs/sdk-overview.mdx`
- `docs/session-persistence.mdx`
- `docs/deploy/local.mdx`
- `docs/deploy/cloudflare.mdx`
- `docs/deploy/vercel.mdx`
- `docs/deploy/daytona.mdx`
- `docs/deploy/e2b.mdx`
- `docs/deploy/docker.mdx`
- `frontend/packages/website/src/components/GetStarted.tsx`
- `.claude/commands/post-release-testing.md`
- `examples/cloudflare/Dockerfile`
- `examples/daytona/src/index.ts`
- `examples/daytona/src/daytona-with-snapshot.ts`
- `examples/docker/src/index.ts`
- `examples/e2b/src/index.ts`
- `examples/vercel/src/index.ts`
- `scripts/release/main.ts`
- `scripts/release/promote-artifacts.ts`
- `scripts/release/sdk.ts`
- `scripts/sandbox-testing/test-sandbox.ts`

3
Cargo.toml
View file

@ -16,6 +16,9 @@ sandbox-agent = { version = "0.2.0", path = "server/packages/sandbox-agent" }
sandbox-agent-error = { version = "0.2.0", path = "server/packages/error" }
sandbox-agent-agent-management = { version = "0.2.0", path = "server/packages/agent-management" }
sandbox-agent-agent-credentials = { version = "0.2.0", path = "server/packages/agent-credentials" }
sandbox-agent-opencode-adapter = { version = "0.2.0", path = "server/packages/opencode-adapter" }
sandbox-agent-opencode-server-manager = { version = "0.2.0", path = "server/packages/opencode-server-manager" }
acp-http-adapter = { version = "0.2.0", path = "server/packages/acp-http-adapter" }
# Serialization
serde = { version = "1.0", features = ["derive"] }

67
docs/advanced/acp-http-client.mdx
View file

@ -1,67 +0,0 @@
---
title: "ACP HTTP Client"
description: "Protocol-pure ACP JSON-RPC over streamable HTTP client."
---
`acp-http-client` is a standalone, low-level package for ACP over HTTP (`/v2/rpc`).
Use it when you want strict ACP protocol behavior with no Sandbox-specific metadata or extension adaptation.
## Install
```bash
npm install acp-http-client
```
## Usage
```ts
import { AcpHttpClient } from "acp-http-client";
const client = new AcpHttpClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
});
await client.initialize({
_meta: {
"sandboxagent.dev": {
agent: "mock",
},
},
});
const session = await client.newSession({
cwd: "/",
mcpServers: [],
_meta: {
"sandboxagent.dev": {
agent: "mock",
},
},
});
const result = await client.prompt({
sessionId: session.sessionId,
prompt: [{ type: "text", text: "hello" }],
});
console.log(result.stopReason);
await client.disconnect();
```
## Scope
- Implements ACP HTTP transport and connection lifecycle.
- Supports ACP requests/notifications and session streaming.
- Does not inject `_meta["sandboxagent.dev"]`.
- Does not wrap `_sandboxagent/*` extension methods/events.
## Transport Contract
- `POST /v2/rpc` is JSON-only. Send `Content-Type: application/json` and `Accept: application/json`.
- `GET /v2/rpc` is SSE-only. Send `Accept: text/event-stream`.
- Keep one active SSE stream per ACP connection id.
- `x-acp-agent` is removed. Provide agent via `_meta["sandboxagent.dev"].agent` on `initialize` and `session/new`.
If you want Sandbox Agent metadata/extensions and higher-level helpers, use `sandbox-agent` and `SandboxAgentClient` instead.

300
docs/agent-sessions.mdx
View file

@ -1,288 +1,90 @@
---
title: "Agent Sessions"
description: "Create sessions and send messages to agents."
description: "Create sessions, prompt agents, and inspect event history."
sidebarTitle: "Sessions"
icon: "comments"
---
Sessions are the unit of interaction with an agent. You create one session per task, then send messages and stream events.
Sessions are the unit of interaction with an agent. Create one session per task, send prompts, and consume event history.
## Session Options
For SDK-based flows, sessions can be restored after runtime/session loss when persistence is enabled.
See [Session Restoration](/session-restoration).
`POST /v1/sessions/{sessionId}` accepts the following fields:
## Create a session
- `agent` (required): `claude`, `codex`, `opencode`, `amp`, or `mock`
- `agentMode`: agent mode string (for example, `build`, `plan`)
- `permissionMode`: permission mode string (`default`, `plan`, `bypass`, etc.)
- `model`: model override (agent-specific)
- `variant`: model variant (agent-specific)
- `agentVersion`: agent version override
- `mcp`: MCP server config map (see `MCP`)
- `skills`: skill path config (see `Skills`)
```ts
import { SandboxAgent } from "sandbox-agent";
## Create A Session
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
await client.createSession("build-session", {
const session = await sdk.createSession({
agent: "codex",
agentMode: "build",
permissionMode: "default",
model: "gpt-4.1",
variant: "reasoning",
agentVersion: "latest",
});
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/build-session" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"agent": "codex",
"agentMode": "build",
"permissionMode": "default",
"model": "gpt-4.1",
"variant": "reasoning",
"agentVersion": "latest"
}'
```
</CodeGroup>
## Send A Message
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
sessionInit: {
cwd: "/",
mcpServers: [],
},
});
await client.postMessage("build-session", {
message: "Summarize the repository structure.",
});
console.log(session.id, session.agentSessionId);
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/build-session/messages" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{"message":"Summarize the repository structure."}'
```
</CodeGroup>
## Send a prompt
## Stream A Turn
```ts
const response = await session.prompt([
{ type: "text", text: "Summarize the repository structure." },
]);
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
const response = await client.postMessageStream("build-session", {
message: "Explain the main entrypoints.",
});
const reader = response.body?.getReader();
if (reader) {
const decoder = new TextDecoder();
while (true) {
const { done, value } = await reader.read();
if (done) break;
console.log(decoder.decode(value, { stream: true }));
}
}
console.log(response.stopReason);
```
```bash cURL
curl -N -X POST "http://127.0.0.1:2468/v1/sessions/build-session/messages/stream" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{"message":"Explain the main entrypoints."}'
```
</CodeGroup>
## Subscribe to live events
## Fetch Events
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
```ts
const unsubscribe = session.onEvent((event) => {
console.log(event.eventIndex, event.sender, event.payload);
});
const events = await client.getEvents("build-session", {
offset: 0,
await session.prompt([
{ type: "text", text: "Explain the main entrypoints." },
]);
unsubscribe();
```
## Fetch persisted event history
```ts
const page = await sdk.getEvents({
sessionId: session.id,
limit: 50,
includeRaw: false,
});
console.log(events.events);
```
```bash cURL
curl -X GET "http://127.0.0.1:2468/v1/sessions/build-session/events?offset=0&limit=50" \
-H "Authorization: Bearer $SANDBOX_TOKEN"
```
</CodeGroup>
`GET /v1/sessions/{sessionId}/get-messages` is an alias for `events`.
## Stream Events (SSE)
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
for await (const event of client.streamEvents("build-session", { offset: 0 })) {
console.log(event.type, event.data);
for (const event of page.items) {
console.log(event.id, event.createdAt, event.sender);
}
```
```bash cURL
curl -N -X GET "http://127.0.0.1:2468/v1/sessions/build-session/events/sse?offset=0" \
-H "Authorization: Bearer $SANDBOX_TOKEN"
```
</CodeGroup>
## List and load sessions
## List Sessions
```ts
const sessions = await sdk.listSessions({ limit: 20 });
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
for (const item of sessions.items) {
console.log(item.id, item.agent, item.createdAt);
}
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
const sessions = await client.listSessions();
console.log(sessions.sessions);
if (sessions.items.length > 0) {
const loaded = await sdk.resumeSession(sessions.items[0]!.id);
await loaded.prompt([{ type: "text", text: "Continue." }]);
}
```
```bash cURL
curl -X GET "http://127.0.0.1:2468/v1/sessions" \
-H "Authorization: Bearer $SANDBOX_TOKEN"
```
</CodeGroup>
## Destroy a session
## Reply To A Question
When the agent asks a question, reply with an array of answers. Each inner array is one multi-select response.
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
await client.replyQuestion("build-session", "question-1", {
answers: [["yes"]],
});
```ts
await sdk.destroySession(session.id);
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/build-session/questions/question-1/reply" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{"answers":[["yes"]]}'
```
</CodeGroup>
## Reject A Question
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
await client.rejectQuestion("build-session", "question-1");
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/build-session/questions/question-1/reject" \
-H "Authorization: Bearer $SANDBOX_TOKEN"
```
</CodeGroup>
## Reply To A Permission Request
Use `once`, `always`, or `reject`.
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
await client.replyPermission("build-session", "permission-1", {
reply: "once",
});
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/build-session/permissions/permission-1/reply" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{"reply":"once"}'
```
</CodeGroup>
## Terminate A Session
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
await client.terminateSession("build-session");
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/build-session/terminate" \
-H "Authorization: Bearer $SANDBOX_TOKEN"
```
</CodeGroup>

64
docs/architecture.mdx Normal file
View file

@ -0,0 +1,64 @@
---
title: "Architecture"
description: "How the client, sandbox, server, and agent fit together."
icon: "microchip"
---
Sandbox Agent runs as an HTTP server inside your sandbox. Your app talks to it remotely.
## Components
- `Your client`: your app code using the `sandbox-agent` SDK.
- `Sandbox`: isolated runtime (E2B, Daytona, Docker, etc.).
- `Sandbox Agent server`: process inside the sandbox exposing HTTP transport.
- `Agent`: Claude/Codex/OpenCode/Amp process managed by Sandbox Agent.
```mermaid placement="top-right"
flowchart LR
CLIENT["Sandbox Agent SDK"]
SERVER["Sandbox Agent server"]
AGENT["Agent process"]
subgraph SANDBOX["Sandbox"]
direction TB
SERVER --> AGENT
end
CLIENT -->|HTTP| SERVER
```
## Suggested Topology
Run the SDK on your backend, then call it from your frontend.
This extra hop is recommended because it keeps auth/token logic on the backend and makes persistence simpler.
```mermaid placement="top-right"
flowchart LR
BROWSER["Browser"]
subgraph BACKEND["Your backend"]
direction TB
SDK["Sandbox Agent SDK"]
end
subgraph SANDBOX_SIMPLE["Sandbox"]
SERVER_SIMPLE["Sandbox Agent server"]
end
BROWSER --> BACKEND
BACKEND --> SDK --> SERVER_SIMPLE
```
### Backend requirements
Your backend layer needs to handle:
- **Long-running connections**: prompts can take minutes.
- **Session affinity**: follow-up messages must reach the same session.
- **State between requests**: session metadata and event history must persist across requests.
- **Graceful recovery**: sessions should resume after backend restarts.
We recommend [Rivet](https://rivet.dev) over serverless because actors natively support the long-lived connections, session routing, and state persistence that agent workloads require.
## Session persistence
For storage driver options and replay behavior, see [Persisting Sessions](/session-persistence).

68
docs/attachments.mdx
View file

@ -1,29 +1,27 @@
---
title: "Attachments"
description: "Upload files into the sandbox and attach them to prompts."
description: "Upload files into the sandbox and reference them in prompts."
sidebarTitle: "Attachments"
icon: "paperclip"
---
Use the filesystem API to upload files, then reference them as attachments when sending prompts.
Use the filesystem API to upload files, then include file references in prompt content.
<Steps>
<Step title="Upload a file">
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
import fs from "node:fs";
const client = new SandboxAgentClient({
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
const buffer = await fs.promises.readFile("./data.csv");
const upload = await client.writeFsFile(
{ path: "./uploads/data.csv", sessionId: "my-session" },
const upload = await sdk.writeFsFile(
{ path: "./uploads/data.csv" },
buffer,
);
@ -31,59 +29,33 @@ Use the filesystem API to upload files, then reference them as attachments when
```
```bash cURL
curl -X PUT "http://127.0.0.1:2468/v1/fs/file?path=./uploads/data.csv&sessionId=my-session" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
curl -X PUT "http://127.0.0.1:2468/v1/fs/file?path=./uploads/data.csv" \
--data-binary @./data.csv
```
</CodeGroup>
The response returns the absolute path that you should use for attachments.
The upload response returns the absolute path.
</Step>
<Step title="Attach the file in a prompt">
<CodeGroup>
<Step title="Reference the file in a prompt">
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const session = await sdk.createSession({ agent: "mock" });
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
await client.postMessage("my-session", {
message: "Please analyze the attached CSV.",
attachments: [
await session.prompt([
{ type: "text", text: "Please analyze the attached CSV." },
{
path: "/home/sandbox/uploads/data.csv",
mime: "text/csv",
filename: "data.csv",
type: "resource_link",
name: "data.csv",
uri: "file:///home/sandbox/uploads/data.csv",
mimeType: "text/csv",
},
],
});
]);
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/my-session/messages" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"message": "Please analyze the attached CSV.",
"attachments": [
{
"path": "/home/sandbox/uploads/data.csv",
"mime": "text/csv",
"filename": "data.csv"
}
]
}'
```
</CodeGroup>
</Step>
</Steps>
## Notes
- Use absolute paths from the upload response to avoid ambiguity.
- If `mime` is omitted, the server defaults to `application/octet-stream`.
- OpenCode receives file parts directly; other agents will see the attachment paths appended to the prompt.
- Use absolute file URIs in `resource_link` blocks.
- If `mimeType` is omitted, the agent/runtime may infer a default.
- Support for non-text resources depends on each agent's ACP prompt capabilities.

370
docs/building-chat-ui.mdx
View file

@ -1,370 +0,0 @@
---
title: "Building a Chat UI"
description: "Build a chat interface using the universal event stream."
icon: "comments"
---
## Setup
### List agents
```ts
const { agents } = await client.listAgents();
// Each agent exposes feature coverage via `capabilities` to determine what UI to show
const claude = agents.find((a) => a.id === "claude");
if (claude?.capabilities.permissions) {
// Show permission approval UI
}
if (claude?.capabilities.questions) {
// Show question response UI
}
```
### Create a session
```ts
const sessionId = `session-${crypto.randomUUID()}`;
await client.createSession(sessionId, {
agent: "claude",
agentMode: "code", // Optional: agent-specific mode
permissionMode: "default", // Optional: "default" | "plan" | "bypass" | "acceptEdits" (Claude: accept edits; Codex: auto-approve file changes; others: default)
model: "claude-sonnet-4", // Optional: model override
});
```
### Send a message
```ts
await client.postMessage(sessionId, { message: "Hello, world!" });
```
### Stream events
Three options for receiving events:
```ts
// Option 1: SSE (recommended for real-time UI)
const stream = client.streamEvents(sessionId, { offset: 0 });
for await (const event of stream) {
handleEvent(event);
}
// Option 2: Polling
const { events, hasMore } = await client.getEvents(sessionId, { offset: 0 });
events.forEach(handleEvent);
// Option 3: Turn streaming (send + stream in one call)
const stream = client.streamTurn(sessionId, { message: "Hello" });
for await (const event of stream) {
handleEvent(event);
}
```
Use `offset` to track the last seen `sequence` number and resume from where you left off.
---
## Handling Events
### Bare minimum
Handle item lifecycle plus turn lifecycle to render a basic chat:
```ts
type ItemState = {
item: UniversalItem;
deltas: string[];
};
const items = new Map<string, ItemState>();
let turnInProgress = false;
function handleEvent(event: UniversalEvent) {
switch (event.type) {
case "turn.started": {
turnInProgress = true;
break;
}
case "turn.ended": {
turnInProgress = false;
break;
}
case "item.started": {
const { item } = event.data as ItemEventData;
items.set(item.item_id, { item, deltas: [] });
break;
}
case "item.delta": {
const { item_id, delta } = event.data as ItemDeltaData;
const state = items.get(item_id);
if (state) {
state.deltas.push(delta);
}
break;
}
case "item.completed": {
const { item } = event.data as ItemEventData;
const state = items.get(item.item_id);
if (state) {
state.item = item;
state.deltas = []; // Clear deltas, use final content
}
break;
}
}
}
```
When rendering:
- Use `turnInProgress` for turn-level UI state (disable send button, show global "Agent is responding", etc.).
- Use `item.status === "in_progress"` for per-item streaming state.
```ts
function renderItem(state: ItemState) {
const { item, deltas } = state;
const isItemLoading = item.status === "in_progress";
// For streaming text, combine item content with accumulated deltas
const text = item.content
.filter((p) => p.type === "text")
.map((p) => p.text)
.join("");
const streamedText = text + deltas.join("");
return {
content: streamedText,
isItemLoading,
isTurnLoading: turnInProgress,
role: item.role,
kind: item.kind,
};
}
```
### Extra events
Handle these for a complete implementation:
```ts
function handleEvent(event: UniversalEvent) {
switch (event.type) {
// ... bare minimum events above ...
case "session.started": {
// Session is ready
break;
}
case "session.ended": {
const { reason, terminated_by } = event.data as SessionEndedData;
// Disable input, show end reason
// reason: "completed" | "error" | "terminated"
// terminated_by: "agent" | "daemon"
break;
}
case "error": {
const { message, code } = event.data as ErrorData;
// Display error to user
break;
}
case "agent.unparsed": {
const { error, location } = event.data as AgentUnparsedData;
// Parsing failure - treat as bug in development
console.error(`Parse error at ${location}: ${error}`);
break;
}
}
}
```
### Content parts
Each item has `content` parts. Render based on `type`:
```ts
function renderContentPart(part: ContentPart) {
switch (part.type) {
case "text":
return <Markdown>{part.text}</Markdown>;
case "tool_call":
return <ToolCall name={part.name} args={part.arguments} />;
case "tool_result":
return <ToolResult output={part.output} />;
case "file_ref":
return <FileChange path={part.path} action={part.action} diff={part.diff} />;
case "reasoning":
return <Reasoning>{part.text}</Reasoning>;
case "status":
return <Status label={part.label} detail={part.detail} />;
case "image":
return <Image src={part.path} />;
}
}
```
---
## Handling Permissions
When `permission.requested` arrives, show an approval UI:
```ts
const pendingPermissions = new Map<string, PermissionEventData>();
function handleEvent(event: UniversalEvent) {
if (event.type === "permission.requested") {
const data = event.data as PermissionEventData;
pendingPermissions.set(data.permission_id, data);
}
if (event.type === "permission.resolved") {
const data = event.data as PermissionEventData;
pendingPermissions.delete(data.permission_id);
}
}
// User clicks approve/deny
async function replyPermission(id: string, reply: "once" | "always" | "reject") {
await client.replyPermission(sessionId, id, { reply });
pendingPermissions.delete(id);
}
```
Render permission requests:
```ts
function PermissionRequest({ data }: { data: PermissionEventData }) {
return (
<div>
<p>Allow: {data.action}</p>
<button onClick={() => replyPermission(data.permission_id, "once")}>
Allow Once
</button>
<button onClick={() => replyPermission(data.permission_id, "always")}>
Always Allow
</button>
<button onClick={() => replyPermission(data.permission_id, "reject")}>
Reject
</button>
</div>
);
}
```
---
## Handling Questions
When `question.requested` arrives, show a selection UI:
```ts
const pendingQuestions = new Map<string, QuestionEventData>();
function handleEvent(event: UniversalEvent) {
if (event.type === "question.requested") {
const data = event.data as QuestionEventData;
pendingQuestions.set(data.question_id, data);
}
if (event.type === "question.resolved") {
const data = event.data as QuestionEventData;
pendingQuestions.delete(data.question_id);
}
}
// User selects answer(s)
async function answerQuestion(id: string, answers: string[][]) {
await client.replyQuestion(sessionId, id, { answers });
pendingQuestions.delete(id);
}
async function rejectQuestion(id: string) {
await client.rejectQuestion(sessionId, id);
pendingQuestions.delete(id);
}
```
Render question requests:
```ts
function QuestionRequest({ data }: { data: QuestionEventData }) {
const [selected, setSelected] = useState<string[]>([]);
return (
<div>
<p>{data.prompt}</p>
{data.options.map((option) => (
<label key={option}>
<input
type="checkbox"
checked={selected.includes(option)}
onChange={(e) => {
if (e.target.checked) {
setSelected([...selected, option]);
} else {
setSelected(selected.filter((s) => s !== option));
}
}}
/>
{option}
</label>
))}
<button onClick={() => answerQuestion(data.question_id, [selected])}>
Submit
</button>
<button onClick={() => rejectQuestion(data.question_id)}>
Reject
</button>
</div>
);
}
```
---
## Testing with Mock Agent
The `mock` agent lets you test UI behaviors without external credentials:
```ts
await client.createSession("test-session", { agent: "mock" });
```
Send `help` to see available commands:
| Command | Tests |
|---------|-------|
| `help` | Lists all commands |
| `demo` | Full UI coverage sequence with markers |
| `markdown` | Streaming markdown rendering |
| `tool` | Tool call + result with file refs |
| `status` | Status item updates |
| `image` | Image content part |
| `permission` | Permission request flow |
| `question` | Question request flow |
| `error` | Error + unparsed events |
| `end` | Session ended event |
| `echo <text>` | Echo text as assistant message |
Any unrecognized text is echoed back as an assistant message.
---
## Reference Implementation
The [Inspector UI](https://github.com/rivet-dev/sandbox-agent/blob/main/frontend/packages/inspector/src/App.tsx)
is a complete reference showing session management, event rendering, and HITL flows.

4
docs/cors.mdx
View file

@ -2,7 +2,6 @@
title: "CORS Configuration"
description: "Configure CORS for browser-based applications."
sidebarTitle: "CORS"
icon: "globe"
---
When calling the Sandbox Agent server from a browser, CORS (Cross-Origin Resource Sharing) controls which origins can make requests.
@ -13,7 +12,6 @@ By default, no CORS origins are allowed. You must explicitly specify origins for
```bash
sandbox-agent server \
--token "$SANDBOX_TOKEN" \
--cors-allow-origin "http://localhost:5173"
```
@ -36,7 +34,6 @@ Specify the flag multiple times to allow multiple origins:
```bash
sandbox-agent server \
--token "$SANDBOX_TOKEN" \
--cors-allow-origin "http://localhost:5173" \
--cors-allow-origin "http://localhost:3000"
```
@ -47,7 +44,6 @@ By default, all methods and headers are allowed. To restrict them:
```bash
sandbox-agent server \
--token "$SANDBOX_TOKEN" \
--cors-allow-origin "https://your-app.com" \
--cors-allow-method "GET" \
--cors-allow-method "POST" \

116
docs/credentials.mdx
View file

@ -1,55 +1,115 @@
---
title: "Credentials"
description: "How sandbox-agent discovers and exposes provider credentials."
icon: "key"
description: "How Sandbox Agent discovers and uses provider credentials."
---
`sandbox-agent` can discover provider credentials from environment variables and local agent config files.
Sandbox Agent discovers API credentials from environment variables and local agent config files.
These credentials are passed through to underlying agent runtimes.
## Supported providers
## Credential sources
- Anthropic
- OpenAI
- Additional provider entries discovered via OpenCode config
Credentials are discovered in priority order.
## Common environment variables
### Environment variables (highest priority)
API keys first:
| Variable | Provider |
| --- | --- |
|----------|----------|
| `ANTHROPIC_API_KEY` | Anthropic |
| `CLAUDE_API_KEY` | Anthropic fallback |
| `OPENAI_API_KEY` | OpenAI |
| `CODEX_API_KEY` | OpenAI fallback |
## Extract credentials (CLI)
OAuth tokens (used when OAuth extraction is enabled):
Show discovered credentials (redacted by default):
| Variable | Provider |
|----------|----------|
| `CLAUDE_CODE_OAUTH_TOKEN` | Anthropic |
| `ANTHROPIC_AUTH_TOKEN` | Anthropic fallback |
```bash
sandbox-agent credentials extract
### Agent config files
| Agent | Config path | Provider |
|-------|-------------|----------|
| Amp | `~/.amp/config.json` | Anthropic |
| Claude Code | `~/.claude.json`, `~/.claude/.credentials.json` | Anthropic |
| Codex | `~/.codex/auth.json` | OpenAI |
| OpenCode | `~/.local/share/opencode/auth.json` | Anthropic/OpenAI |
## Provider requirements by agent
| Agent | Required provider |
|-------|-------------------|
| Claude Code | Anthropic |
| Amp | Anthropic |
| Codex | OpenAI |
| OpenCode | Anthropic or OpenAI |
| Mock | None |
## Error handling behavior
Credential extraction is best-effort:
- Missing or malformed files are skipped.
- Discovery continues to later sources.
- Missing credentials mark providers unavailable instead of failing server startup.
When prompting, Sandbox Agent does not pre-validate provider credentials. Agent-native authentication errors surface through session events/output.
## Checking credential status
### API
`GET /v1/agents` includes `credentialsAvailable` per agent.
```json
{
"agents": [
{
"id": "claude",
"installed": true,
"credentialsAvailable": true
},
{
"id": "codex",
"installed": true,
"credentialsAvailable": false
}
]
}
```
Reveal raw values:
### TypeScript SDK
```bash
sandbox-agent credentials extract --reveal
```typescript
const result = await sdk.listAgents();
for (const agent of result.agents) {
console.log(`${agent.id}: ${agent.credentialsAvailable ? "authenticated" : "no credentials"}`);
}
```
Filter by agent/provider:
## Passing credentials explicitly
Set environment variables before starting Sandbox Agent:
```bash
sandbox-agent credentials extract --agent codex
sandbox-agent credentials extract --provider openai
export ANTHROPIC_API_KEY=sk-ant-...
export OPENAI_API_KEY=sk-...
sandbox-agent daemon start
```
Emit shell exports:
Or with SDK-managed local spawn:
```bash
sandbox-agent credentials extract-env --export
```typescript
import { SandboxAgent } from "sandbox-agent";
const sdk = await SandboxAgent.start({
spawn: {
env: {
ANTHROPIC_API_KEY: process.env.MY_ANTHROPIC_KEY,
},
},
});
```
## Notes
- Discovery is best-effort: missing/invalid files do not crash extraction.
- v2 does not expose legacy v1 `credentialsAvailable` agent fields.
- Authentication failures are surfaced by the selected ACP agent process/agent during ACP requests.

204
docs/custom-tools.mdx
View file

@ -5,132 +5,97 @@ sidebarTitle: "Custom Tools"
icon: "wrench"
---
There are two ways to give agents custom tools that run inside the sandbox:
There are two common patterns for sandbox-local custom tooling:
| | MCP Server | Skill |
|---|---|---|
| **How it works** | Sandbox Agent spawns your MCP server process and routes tool calls to it via stdio | A markdown file that instructs the agent to run your script with `node` (or any command) |
| **Tool discovery** | Agent sees tools automatically via MCP protocol | Agent reads instructions from the skill file |
| **Best for** | Structured tools with typed inputs/outputs | Lightweight scripts with natural-language instructions |
| **Requires** | `@modelcontextprotocol/sdk` dependency | Just a markdown file and a script |
| **How it works** | Agent connects to an MCP server (`mcpServers`) | Agent follows `SKILL.md` instructions and runs scripts |
| **Best for** | Typed tool calls and structured protocols | Lightweight task-specific guidance |
| **Requires** | MCP server process (stdio/http/sse) | Script + `SKILL.md` |
Both approaches execute code inside the sandbox, so your tools have full access to the sandbox filesystem, network, and installed system tools.
## Option A: Tools via MCP
## Option A: MCP server (stdio)
<Steps>
<Step title="Write your MCP server">
Create an MCP server that exposes tools using `@modelcontextprotocol/sdk` with `StdioServerTransport`. This server will run inside the sandbox.
<Step title="Write and bundle your MCP server">
```ts src/mcp-server.ts
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import { z } from "zod";
const server = new McpServer({
name: "rand",
version: "1.0.0",
});
const server = new McpServer({ name: "rand", version: "1.0.0" });
server.tool(
"random_number",
"Generate a random integer between min and max (inclusive)",
"Generate a random integer between min and max",
{
min: z.number().describe("Minimum value"),
max: z.number().describe("Maximum value"),
min: z.number(),
max: z.number(),
},
async ({ min, max }) => ({
content: [{ type: "text", text: String(Math.floor(Math.random() * (max - min + 1)) + min) }],
}),
);
const transport = new StdioServerTransport();
await server.connect(transport);
await server.connect(new StdioServerTransport());
```
This is a simple example. Your MCP server runs inside the sandbox, so you can execute any code you'd like: query databases, call internal APIs, run shell commands, or interact with any service available in the container.
</Step>
<Step title="Package the MCP server">
Bundle into a single JS file so it can be uploaded and executed without a `node_modules` folder.
```bash
npx esbuild src/mcp-server.ts --bundle --format=cjs --platform=node --target=node18 --minify --outfile=dist/mcp-server.cjs
npx esbuild src/mcp-server.ts --bundle --format=cjs --platform=node --target=node18 --outfile=dist/mcp-server.cjs
```
This creates `dist/mcp-server.cjs` ready to upload.
</Step>
<Step title="Create sandbox and upload MCP server">
Start your sandbox, then write the bundled file into it.
<Step title="Upload it into the sandbox">
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
```ts
import { SandboxAgent } from "sandbox-agent";
import fs from "node:fs";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
const sdk = await SandboxAgent.connect({ baseUrl: "http://127.0.0.1:2468" });
const content = await fs.promises.readFile("./dist/mcp-server.cjs");
await client.writeFsFile(
{ path: "/opt/mcp/custom-tools/mcp-server.cjs" },
content,
);
await sdk.writeFsFile({ path: "/opt/mcp/custom-tools/mcp-server.cjs" }, content);
```
```bash cURL
```bash
curl -X PUT "http://127.0.0.1:2468/v1/fs/file?path=/opt/mcp/custom-tools/mcp-server.cjs" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
--data-binary @./dist/mcp-server.cjs
```
</CodeGroup>
</Step>
<Step title="Create a session">
Point an MCP server config at the bundled JS file. When the session starts, Sandbox Agent spawns the MCP server process and routes tool calls to it.
<Step title="Register MCP config and create a session">
<CodeGroup>
```ts TypeScript
await client.createSession("custom-tools", {
agent: "claude",
mcp: {
customTools: {
type: "local",
command: ["node", "/opt/mcp/custom-tools/mcp-server.cjs"],
```ts
await sdk.setMcpConfig(
{
directory: "/workspace",
mcpName: "customTools",
},
{
type: "local",
command: "node",
args: ["/opt/mcp/custom-tools/mcp-server.cjs"],
},
);
const session = await sdk.createSession({
agent: "claude",
sessionInit: {
cwd: "/workspace",
},
});
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/custom-tools" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"agent": "claude",
"mcp": {
"customTools": {
"type": "local",
"command": ["node", "/opt/mcp/custom-tools/mcp-server.cjs"]
}
}
}'
await session.prompt([
{ type: "text", text: "Use the random_number tool with min=1 and max=10." },
]);
```
</CodeGroup>
</Step>
</Steps>
## Option B: Tools via Skills
Skills are markdown files that instruct the agent how to use a script. Upload the script and a skill file, then point the session at the skill directory.
## Option B: Skills
<Steps>
<Step title="Write your script">
Write a script that the agent will execute. This runs inside the sandbox just like an MCP server, but the agent invokes it directly via its shell tool.
<Step title="Write script + skill file">
```ts src/random-number.ts
const min = Number(process.argv[2]);
@ -143,105 +108,56 @@ Skills are markdown files that instruct the agent how to use a script. Upload th
console.log(Math.floor(Math.random() * (max - min + 1)) + min);
```
</Step>
<Step title="Write a skill file">
Create a `SKILL.md` that tells the agent what the script does and how to run it. The frontmatter `name` and `description` fields are required. See [Skill authoring best practices](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices) for tips on writing effective skills.
```md SKILL.md
````md SKILL.md
---
name: random-number
description: Generate a random integer between min and max (inclusive). Use when the user asks for a random number.
description: Generate a random integer between min and max.
---
To generate a random number, run:
Run:
```bash
node /opt/skills/random-number/random-number.cjs <min> <max>
```
This prints a single random integer between min and max (inclusive).
</Step>
<Step title="Package the script">
Bundle the script just like an MCP server so it has no dependencies at runtime.
````
```bash
npx esbuild src/random-number.ts --bundle --format=cjs --platform=node --target=node18 --minify --outfile=dist/random-number.cjs
npx esbuild src/random-number.ts --bundle --format=cjs --platform=node --target=node18 --outfile=dist/random-number.cjs
```
</Step>
<Step title="Create sandbox and upload files">
Upload both the bundled script and the skill file.
<Step title="Upload files">
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
```ts
import fs from "node:fs";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
const script = await fs.promises.readFile("./dist/random-number.cjs");
await client.writeFsFile(
{ path: "/opt/skills/random-number/random-number.cjs" },
script,
);
await sdk.writeFsFile({ path: "/opt/skills/random-number/random-number.cjs" }, script);
const skill = await fs.promises.readFile("./SKILL.md");
await client.writeFsFile(
{ path: "/opt/skills/random-number/SKILL.md" },
skill,
);
await sdk.writeFsFile({ path: "/opt/skills/random-number/SKILL.md" }, skill);
```
```bash cURL
curl -X PUT "http://127.0.0.1:2468/v1/fs/file?path=/opt/skills/random-number/random-number.cjs" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
--data-binary @./dist/random-number.cjs
curl -X PUT "http://127.0.0.1:2468/v1/fs/file?path=/opt/skills/random-number/SKILL.md" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
--data-binary @./SKILL.md
```
</CodeGroup>
</Step>
<Step title="Create a session">
Point the session at the skill directory. The agent reads `SKILL.md` and learns how to use your script.
<Step title="Use in a session">
<CodeGroup>
```ts TypeScript
await client.createSession("custom-tools", {
```ts
const session = await sdk.createSession({
agent: "claude",
skills: {
sources: [
{ type: "local", source: "/opt/skills/random-number" },
],
sessionInit: {
cwd: "/workspace",
},
});
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/custom-tools" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"agent": "claude",
"skills": {
"sources": [
{ "type": "local", "source": "/opt/skills/random-number" }
]
}
}'
await session.prompt([
{ type: "text", text: "Use the random-number skill to pick a number from 1 to 100." },
]);
```
</CodeGroup>
</Step>
</Steps>
## Notes
- The sandbox image must include a Node.js runtime that can execute the bundled files.
- The sandbox runtime must include Node.js (or your chosen runtime).
- For persistent skill-source wiring by directory, see [Skills](/skills-config).

71
docs/daemon.mdx
View file

@ -1,96 +1,69 @@
---
title: "Daemon"
description: "Background daemon lifecycle, auto-upgrade, and management."
icon: "microchip"
description: "Background daemon lifecycle and management."
---
The sandbox-agent daemon is a background server process that stays running between sessions. Commands like `sandbox-agent opencode` and `gigacode` automatically start it when needed and restart it when the binary is updated.
The sandbox-agent daemon is a background server process. Commands like `sandbox-agent opencode` and `gigacode` can ensure it is running.
## How it works
1. When you run `sandbox-agent opencode`, `sandbox-agent daemon start`, or `gigacode`, the CLI checks if a daemon is already healthy at the configured host and port.
2. If no daemon is running, one is spawned in the background with stdout/stderr redirected to a log file.
3. The CLI writes a PID file and a build ID file to track the running process and its version.
4. On subsequent invocations, if the daemon is still running but was built from a different commit, the CLI automatically stops the old daemon and starts a new one.
1. A daemon-aware command checks for a healthy daemon at host/port.
2. If missing, it starts one in the background and records PID/version files.
3. Subsequent checks can compare build/version and restart when required.
## Auto-upgrade
## Auto-upgrade behavior
Each build of sandbox-agent embeds a unique **build ID** (the git short hash, or a version-timestamp fallback). When a daemon is started, this build ID is written to a version file alongside the PID file.
On every invocation of `ensure_running` (called by `opencode`, `gigacode`, and `daemon start`), the CLI compares the stored build ID against the current binary's build ID. If they differ, the running daemon is stopped and replaced automatically:
```
daemon outdated (build a1b2c3d -> f4e5d6c), restarting...
```
This means installing a new version of sandbox-agent and running any daemon-aware command is enough to upgrade — no manual restart needed.
- `sandbox-agent opencode` and `gigacode` use ensure-running behavior with upgrade checks.
- `sandbox-agent daemon start` uses direct start by default.
- `sandbox-agent daemon start --upgrade` uses ensure-running behavior (including version check/restart).
## Managing the daemon
### Start
Start a daemon in the background. If one is already running and healthy, this is a no-op.
```bash
sandbox-agent daemon start [OPTIONS]
```
| Option | Default | Description |
|--------|---------|-------------|
| `-H, --host <HOST>` | `127.0.0.1` | Host to bind to |
| `-p, --port <PORT>` | `2468` | Port to bind to |
| `-t, --token <TOKEN>` | - | Authentication token |
| `-n, --no-token` | - | Disable authentication |
| `-H, --host <HOST>` | `127.0.0.1` | Host |
| `-p, --port <PORT>` | `2468` | Port |
| `--upgrade` | false | Use ensure-running + upgrade behavior |
```bash
sandbox-agent daemon start --no-token
sandbox-agent daemon start
sandbox-agent daemon start --upgrade
```
### Stop
Stop a running daemon. Sends SIGTERM and waits up to 5 seconds for a graceful shutdown before falling back to SIGKILL.
```bash
sandbox-agent daemon stop [OPTIONS]
```
| Option | Default | Description |
|--------|---------|-------------|
| `-H, --host <HOST>` | `127.0.0.1` | Host of the daemon |
| `-p, --port <PORT>` | `2468` | Port of the daemon |
```bash
sandbox-agent daemon stop
```
| `-H, --host <HOST>` | `127.0.0.1` | Host |
| `-p, --port <PORT>` | `2468` | Port |
### Status
Show whether the daemon is running, its PID, build ID, and log path.
```bash
sandbox-agent daemon status [OPTIONS]
```
| Option | Default | Description |
|--------|---------|-------------|
| `-H, --host <HOST>` | `127.0.0.1` | Host of the daemon |
| `-p, --port <PORT>` | `2468` | Port of the daemon |
```bash
sandbox-agent daemon status
# Daemon running (PID 12345, build a1b2c3d, logs: ~/.local/share/sandbox-agent/daemon/daemon-127-0-0-1-2468.log)
```
If the daemon was started with an older binary, the status includes an `[outdated, restart recommended]` notice.
| `-H, --host <HOST>` | `127.0.0.1` | Host |
| `-p, --port <PORT>` | `2468` | Port |
## Files
All daemon state files live under the sandbox-agent data directory (typically `~/.local/share/sandbox-agent/daemon/`):
Daemon state is stored under the sandbox-agent data directory (for example `~/.local/share/sandbox-agent/daemon/`):
| File | Purpose |
|------|---------|
| `daemon-{host}-{port}.pid` | PID of the running daemon process |
| `daemon-{host}-{port}.version` | Build ID of the running daemon |
| `daemon-{host}-{port}.log` | Daemon stdout/stderr log output |
Multiple daemons can run on different host/port combinations without conflicting.
| `daemon-{host}-{port}.pid` | PID of running daemon |
| `daemon-{host}-{port}.version` | Build/version marker |
| `daemon-{host}-{port}.log` | Daemon stdout/stderr log |

191
docs/deploy/cloudflare.mdx
View file

@ -1,21 +1,19 @@
---
title: "Cloudflare"
description: "Deploy the daemon inside a Cloudflare Sandbox."
description: "Deploy Sandbox Agent inside a Cloudflare Sandbox."
---
## Prerequisites
- Cloudflare account with Workers Paid plan
- Docker running locally for `wrangler dev`
- `ANTHROPIC_API_KEY` or `OPENAI_API_KEY` for the coding agents
- Cloudflare account with Workers paid plan
- Docker for local `wrangler dev`
- `ANTHROPIC_API_KEY` or `OPENAI_API_KEY`
<Note>
Cloudflare Sandbox SDK is in beta. See [Sandbox SDK docs](https://developers.cloudflare.com/sandbox/) for details.
Cloudflare Sandbox SDK is beta. See [Sandbox SDK docs](https://developers.cloudflare.com/sandbox/).
</Note>
## Quick Start
Create a new Sandbox SDK project:
## Quick start
```bash
npm create cloudflare@latest -- my-sandbox --template=cloudflare/sandbox-sdk/examples/minimal
@ -24,64 +22,16 @@ cd my-sandbox
## Dockerfile
Create a `Dockerfile` with sandbox-agent and agents pre-installed:
```dockerfile
FROM cloudflare/sandbox:0.7.0
# Install sandbox-agent
RUN curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh
RUN curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh
RUN sandbox-agent install-agent claude && sandbox-agent install-agent codex
# Pre-install agents
RUN sandbox-agent install-agent claude && \
sandbox-agent install-agent codex
# Required for local development with wrangler dev
EXPOSE 8000
```
<Note>
The `EXPOSE 8000` directive is required for `wrangler dev` to proxy requests to the container. Port 3000 is reserved for the Cloudflare control plane.
</Note>
## Wrangler Configuration
Update `wrangler.jsonc` to use your Dockerfile:
```jsonc
{
"name": "my-sandbox-agent",
"main": "src/index.ts",
"compatibility_date": "2025-01-01",
"compatibility_flags": ["nodejs_compat"],
"containers": [
{
"class_name": "Sandbox",
"image": "./Dockerfile",
"instance_type": "lite",
"max_instances": 1
}
],
"durable_objects": {
"bindings": [
{
"class_name": "Sandbox",
"name": "Sandbox"
}
]
},
"migrations": [
{
"new_sqlite_classes": ["Sandbox"],
"tag": "v1"
}
]
}
```
## TypeScript Example
This example proxies requests to sandbox-agent via `containerFetch`, which works reliably in both local development and production:
## TypeScript proxy example
```typescript
import { getSandbox, type Sandbox } from "@cloudflare/sandbox";
@ -95,158 +45,87 @@ type Env = {
const PORT = 8000;
/** Check if sandbox-agent is already running */
async function isServerRunning(sandbox: Sandbox): Promise<boolean> {
try {
const result = await sandbox.exec(`curl -sf http://localhost:${PORT}/v2/health`);
const result = await sandbox.exec(`curl -sf http://localhost:${PORT}/v1/health`);
return result.success;
} catch {
return false;
}
}
/** Ensure sandbox-agent is running in the container */
async function ensureRunning(sandbox: Sandbox, env: Env): Promise<void> {
if (await isServerRunning(sandbox)) return;
// Set environment variables for agents
const envVars: Record<string, string> = {};
if (env.ANTHROPIC_API_KEY) envVars.ANTHROPIC_API_KEY = env.ANTHROPIC_API_KEY;
if (env.OPENAI_API_KEY) envVars.OPENAI_API_KEY = env.OPENAI_API_KEY;
await sandbox.setEnvVars(envVars);
// Start sandbox-agent server
await sandbox.startProcess(
`sandbox-agent server --no-token --host 0.0.0.0 --port ${PORT}`
);
await sandbox.startProcess(`sandbox-agent server --no-token --host 0.0.0.0 --port ${PORT}`);
// Poll health endpoint until server is ready
for (let i = 0; i < 30; i++) {
if (await isServerRunning(sandbox)) return;
await new Promise((r) => setTimeout(r, 200));
}
throw new Error("sandbox-agent failed to start");
}
export default {
async fetch(request: Request, env: Env): Promise<Response> {
const url = new URL(request.url);
// Proxy requests: /sandbox/:name/v2/...
const match = url.pathname.match(/^\/sandbox\/([^/]+)(\/.*)?$/);
if (match) {
const [, name, path = "/"] = match;
const sandbox = getSandbox(env.Sandbox, name);
await ensureRunning(sandbox, env);
// Proxy request to container
return sandbox.containerFetch(
new Request(`http://localhost${path}${url.search}`, request),
PORT
);
if (!match) {
return new Response("Not found", { status: 404 });
}
return new Response("Not found", { status: 404 });
const [, name, path = "/"] = match;
const sandbox = getSandbox(env.Sandbox, name);
await ensureRunning(sandbox, env);
return sandbox.containerFetch(
new Request(`http://localhost${path}${url.search}`, request),
PORT,
);
},
};
```
## Connect from Client
## Connect from a client
```typescript
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
// Connect via the proxy endpoint
const client = new SandboxAgentClient({
const sdk = await SandboxAgent.connect({
baseUrl: "http://localhost:8787/sandbox/my-sandbox",
agent: "mock",
});
// Wait for server to be ready
for (let i = 0; i < 30; i++) {
try {
await client.getHealth();
break;
} catch {
await new Promise((r) => setTimeout(r, 1000));
}
}
const session = await sdk.createSession({ agent: "claude" });
// Create a session and start coding
await client.createSession("my-session", { agent: "claude" });
await client.postMessage("my-session", {
message: "Summarize this repository",
const off = session.onEvent((event) => {
console.log(event.sender, event.payload);
});
for await (const event of client.streamEvents("my-session")) {
// Auto-approve permissions
if (event.type === "permission.requested") {
await client.replyPermission("my-session", event.data.permission_id, {
reply: "once",
});
}
// Handle text output
if (event.type === "item.delta" && event.data?.delta) {
process.stdout.write(event.data.delta);
}
}
await session.prompt([{ type: "text", text: "Summarize this repository" }]);
off();
```
## Environment Variables
Use `.dev.vars` for local development:
```bash
echo "ANTHROPIC_API_KEY=your-api-key" > .dev.vars
```
<Warning>
Use plain `KEY=value` format in `.dev.vars`. Do not use `export KEY=value` - wrangler won't parse the bash syntax.
</Warning>
<Note>
The `.dev.vars` file is automatically gitignored and only used during local development with `npm run dev`.
</Note>
For production, set secrets via wrangler:
```bash
wrangler secret put ANTHROPIC_API_KEY
```
## Local Development
Start the development server:
## Local development
```bash
npm run dev
```
<Note>
First run builds the Docker container (2-3 minutes). Subsequent runs are much faster.
</Note>
Test with curl:
Test health:
```bash
curl http://localhost:8787/sandbox/demo/v2/health
curl http://localhost:8787/sandbox/demo/v1/health
```
<Tip>
Containers cache environment variables. If you change `.dev.vars`, either use a new sandbox name or clear existing containers:
```bash
docker ps -a | grep sandbox | awk '{print $1}' | xargs -r docker rm -f
```
</Tip>
## Production Deployment
Deploy to Cloudflare:
## Production deployment
```bash
wrangler deploy
```
For production with preview URLs (direct container access), you'll need a custom domain with wildcard DNS routing. See [Cloudflare Production Deployment](https://developers.cloudflare.com/sandbox/guides/production-deployment/) for setup instructions.

46
docs/deploy/daytona.mdx
View file

@ -1,63 +1,52 @@
---
title: "Daytona"
description: "Run the daemon in a Daytona workspace."
description: "Run Sandbox Agent in a Daytona workspace."
---
<Warning>
Daytona Tier 3+ is required to access api.anthropic.com and api.openai.com. Tier 1/2 sandboxes have restricted network access that will cause agent failures. See [Daytona network limits](https://www.daytona.io/docs/en/network-limits/) for details.
Daytona Tier 3+ is required for access to common model provider endpoints.
See [Daytona network limits](https://www.daytona.io/docs/en/network-limits/).
</Warning>
## Prerequisites
- `DAYTONA_API_KEY` environment variable
- `ANTHROPIC_API_KEY` or `OPENAI_API_KEY` for the coding agents
- `DAYTONA_API_KEY`
- `ANTHROPIC_API_KEY` or `OPENAI_API_KEY`
## TypeScript Example
## TypeScript example
```typescript
import { Daytona } from "@daytonaio/sdk";
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
const daytona = new Daytona();
// Pass API keys to the sandbox
const envVars: Record<string, string> = {};
if (process.env.ANTHROPIC_API_KEY) envVars.ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;
if (process.env.OPENAI_API_KEY) envVars.OPENAI_API_KEY = process.env.OPENAI_API_KEY;
const sandbox = await daytona.create({ envVars });
// Install sandbox-agent
await sandbox.process.executeCommand(
"curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh"
"curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh"
);
// Start the server in the background
await sandbox.process.executeCommand(
"nohup sandbox-agent server --no-token --host 0.0.0.0 --port 3000 >/tmp/sandbox-agent.log 2>&1 &"
);
// Wait for server to be ready
await new Promise((r) => setTimeout(r, 2000));
// Get the public URL
const baseUrl = (await sandbox.getSignedPreviewUrl(3000, 4 * 60 * 60)).url;
const sdk = await SandboxAgent.connect({ baseUrl });
// Connect and use the SDK
const client = new SandboxAgentClient({ baseUrl, agent: "mock" });
const session = await sdk.createSession({ agent: "claude" });
await session.prompt([{ type: "text", text: "Summarize this repository" }]);
await client.createSession("my-session", {
agent: "claude",
permissionMode: "default",
});
// Cleanup when done
await sandbox.delete();
```
## Using Snapshots for Faster Startup
For production, use snapshots with pre-installed binaries:
## Using snapshots for faster startup
```typescript
import { Daytona, Image } from "@daytonaio/sdk";
@ -65,7 +54,6 @@ import { Daytona, Image } from "@daytonaio/sdk";
const daytona = new Daytona();
const SNAPSHOT = "sandbox-agent-ready";
// Create snapshot once (takes 2-3 minutes)
const hasSnapshot = await daytona.snapshot.get(SNAPSHOT).then(() => true, () => false);
if (!hasSnapshot) {
@ -73,18 +61,10 @@ if (!hasSnapshot) {
name: SNAPSHOT,
image: Image.base("ubuntu:22.04").runCommands(
"apt-get update && apt-get install -y curl ca-certificates",
"curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh",
"curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh",
"sandbox-agent install-agent claude",
"sandbox-agent install-agent codex",
),
});
}
// Now sandboxes start instantly
const sandbox = await daytona.create({
snapshot: SNAPSHOT,
envVars,
});
```
See [Daytona Snapshots](https://daytona.io/docs/snapshots) for details.

34
docs/deploy/docker.mdx
View file

@ -1,15 +1,15 @@
---
title: "Docker"
description: "Build and run the daemon in a Docker container."
description: "Build and run Sandbox Agent in a Docker container."
---
<Warning>
Docker is not recommended for production. Standard Docker containers don't provide sufficient isolation for running untrusted code. Use a dedicated sandbox provider like E2B or Daytona for production workloads.
Docker is not recommended for production isolation of untrusted workloads. Use dedicated sandbox providers (E2B, Daytona, etc.) for stronger isolation.
</Warning>
## Quick Start
## Quick start
Run sandbox-agent in a container with agents pre-installed:
Run Sandbox Agent with agents pre-installed:
```bash
docker run --rm -p 3000:3000 \
@ -17,23 +17,21 @@ docker run --rm -p 3000:3000 \
-e OPENAI_API_KEY="$OPENAI_API_KEY" \
alpine:latest sh -c "\
apk add --no-cache curl ca-certificates libstdc++ libgcc bash && \
curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh && \
curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh && \
sandbox-agent install-agent claude && \
sandbox-agent install-agent codex && \
sandbox-agent server --no-token --host 0.0.0.0 --port 3000"
```
<Note>
Alpine is required because Claude Code is built for musl libc. Debian/Ubuntu images use glibc and won't work.
Alpine is required for some agent binaries that target musl libc.
</Note>
Access the API at `http://localhost:3000`.
## TypeScript with dockerode
```typescript
import Docker from "dockerode";
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
const docker = new Docker();
const PORT = 3000;
@ -42,7 +40,7 @@ const container = await docker.createContainer({
Image: "alpine:latest",
Cmd: ["sh", "-c", [
"apk add --no-cache curl ca-certificates libstdc++ libgcc bash",
"curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh",
"curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh",
"sandbox-agent install-agent claude",
"sandbox-agent install-agent codex",
`sandbox-agent server --no-token --host 0.0.0.0 --port ${PORT}`,
@ -60,24 +58,18 @@ const container = await docker.createContainer({
await container.start();
// Wait for server and connect
const baseUrl = `http://127.0.0.1:${PORT}`;
const client = new SandboxAgentClient({ baseUrl, agent: "mock" });
const sdk = await SandboxAgent.connect({ baseUrl });
// Use the client...
await client.createSession("my-session", {
agent: "claude",
permissionMode: "default",
});
const session = await sdk.createSession({ agent: "claude" });
await session.prompt([{ type: "text", text: "Summarize this repository." }]);
```
## Building from Source
To build a static binary for use in minimal containers:
## Building from source
```bash
docker build -f docker/release/linux-x86_64.Dockerfile -t sandbox-agent-build .
docker run --rm -v "$PWD/artifacts:/artifacts" sandbox-agent-build
```
The binary will be at `./artifacts/sandbox-agent-x86_64-unknown-linux-musl`.
Binary output: `./artifacts/sandbox-agent-x86_64-unknown-linux-musl`.

57
docs/deploy/e2b.mdx
View file

@ -1,79 +1,52 @@
---
title: "E2B"
description: "Deploy the daemon inside an E2B sandbox."
description: "Deploy Sandbox Agent inside an E2B sandbox."
---
## Prerequisites
- `E2B_API_KEY` environment variable
- `ANTHROPIC_API_KEY` or `OPENAI_API_KEY` for the coding agents
- `E2B_API_KEY`
- `ANTHROPIC_API_KEY` or `OPENAI_API_KEY`
## TypeScript Example
## TypeScript example
```typescript
import { Sandbox } from "@e2b/code-interpreter";
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
// Pass API keys to the sandbox
const envs: Record<string, string> = {};
if (process.env.ANTHROPIC_API_KEY) envs.ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;
if (process.env.OPENAI_API_KEY) envs.OPENAI_API_KEY = process.env.OPENAI_API_KEY;
const sandbox = await Sandbox.create({ allowInternetAccess: true, envs });
// Install sandbox-agent
await sandbox.commands.run(
"curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh"
"curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh"
);
// Install agents before starting the server
await sandbox.commands.run("sandbox-agent install-agent claude");
await sandbox.commands.run("sandbox-agent install-agent codex");
// Start the server in the background
await sandbox.commands.run(
"sandbox-agent server --no-token --host 0.0.0.0 --port 3000",
{ background: true }
);
// Connect to the server
const baseUrl = `https://${sandbox.getHost(3000)}`;
const client = new SandboxAgentClient({ baseUrl, agent: "mock" });
const sdk = await SandboxAgent.connect({ baseUrl });
// Wait for server to be ready
for (let i = 0; i < 30; i++) {
try {
await client.getHealth();
break;
} catch {
await new Promise((r) => setTimeout(r, 1000));
}
}
// Create a session and start coding
await client.createSession("my-session", {
agent: "claude",
permissionMode: "default",
const session = await sdk.createSession({ agent: "claude" });
const off = session.onEvent((event) => {
console.log(event.sender, event.payload);
});
await client.postMessage("my-session", {
message: "Summarize this repository",
});
await session.prompt([{ type: "text", text: "Summarize this repository" }]);
off();
for await (const event of client.streamEvents("my-session")) {
console.log(event.type, event.data);
}
// Cleanup
await sandbox.kill();
```
## Faster Cold Starts
## Faster cold starts
For faster startup, create a custom E2B template with sandbox-agent and agents pre-installed:
1. Create a template with the install script baked in
2. Pre-install agents: `sandbox-agent install-agent claude codex`
3. Use the template ID when creating sandboxes
See [E2B Custom Templates](https://e2b.dev/docs/sandbox-template) for details.
For faster startup, create a custom E2B template with Sandbox Agent and target agents pre-installed.
See [E2B Custom Templates](https://e2b.dev/docs/sandbox-template).

29
docs/deploy/local.mdx
View file

@ -1,52 +1,53 @@
---
title: "Local"
description: "Run the daemon locally for development."
description: "Run Sandbox Agent locally for development."
---
For local development, you can run the daemon directly on your machine.
For local development, run Sandbox Agent directly on your machine.
## With the CLI
```bash
# Install
curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh
curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh
# Run
sandbox-agent server --no-token --host 127.0.0.1 --port 2468
```
Or with npm or Bun:
Or with npm/Bun:
<Tabs>
<Tab title="npx">
```bash
npx sandbox-agent server --no-token --host 127.0.0.1 --port 2468
npx @sandbox-agent/cli@0.2.x server --no-token --host 127.0.0.1 --port 2468
```
</Tab>
<Tab title="bunx">
```bash
bunx sandbox-agent server --no-token --host 127.0.0.1 --port 2468
bunx @sandbox-agent/cli@0.2.x server --no-token --host 127.0.0.1 --port 2468
```
</Tab>
</Tabs>
## With the TypeScript SDK
The SDK can automatically spawn and manage the server as a subprocess:
The SDK can spawn and manage the server as a subprocess:
```typescript
import { SandboxAgent } from "sandbox-agent";
// Spawns sandbox-agent server as a subprocess
const client = await SandboxAgent.start();
const sdk = await SandboxAgent.start();
await client.createSession("my-session", {
const session = await sdk.createSession({
agent: "claude",
permissionMode: "default",
});
// When done
await client.dispose();
await session.prompt([
{ type: "text", text: "Summarize this repository." },
]);
await sdk.dispose();
```
This installs the binary (if needed) and starts the server on a random available port. No manual setup required.
This starts the server on an available local port and connects automatically.

57
docs/deploy/vercel.mdx
View file

@ -1,47 +1,39 @@
---
title: "Vercel"
description: "Deploy the daemon inside a Vercel Sandbox."
description: "Deploy Sandbox Agent inside a Vercel Sandbox."
---
## Prerequisites
- `VERCEL_OIDC_TOKEN` or `VERCEL_ACCESS_TOKEN` environment variable
- `ANTHROPIC_API_KEY` or `OPENAI_API_KEY` for the coding agents
- `VERCEL_OIDC_TOKEN` or `VERCEL_ACCESS_TOKEN`
- `ANTHROPIC_API_KEY` or `OPENAI_API_KEY`
## TypeScript Example
## TypeScript example
```typescript
import { Sandbox } from "@vercel/sandbox";
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
// Pass API keys to the sandbox
const envs: Record<string, string> = {};
if (process.env.ANTHROPIC_API_KEY) envs.ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;
if (process.env.OPENAI_API_KEY) envs.OPENAI_API_KEY = process.env.OPENAI_API_KEY;
// Create sandbox with port 3000 exposed
const sandbox = await Sandbox.create({
runtime: "node24",
ports: [3000],
});
// Helper to run commands
const run = async (cmd: string, args: string[] = []) => {
const result = await sandbox.runCommand({ cmd, args, env: envs });
if (result.exitCode !== 0) {
throw new Error(`Command failed: ${cmd} ${args.join(" ")}`);
}
return result;
};
// Install sandbox-agent
await run("sh", ["-c", "curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh"]);
// Install agents before starting the server
await run("sh", ["-c", "curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh"]);
await run("sandbox-agent", ["install-agent", "claude"]);
await run("sandbox-agent", ["install-agent", "codex"]);
// Start the server in the background
await sandbox.runCommand({
cmd: "sandbox-agent",
args: ["server", "--no-token", "--host", "0.0.0.0", "--port", "3000"],
@ -49,43 +41,22 @@ await sandbox.runCommand({
detached: true,
});
// Connect to the server
const baseUrl = sandbox.domain(3000);
const client = new SandboxAgentClient({ baseUrl, agent: "mock" });
const sdk = await SandboxAgent.connect({ baseUrl });
// Wait for server to be ready
for (let i = 0; i < 30; i++) {
try {
await client.getHealth();
break;
} catch {
await new Promise((r) => setTimeout(r, 1000));
}
}
const session = await sdk.createSession({ agent: "claude" });
// Create a session and start coding
await client.createSession("my-session", {
agent: "claude",
permissionMode: "default",
const off = session.onEvent((event) => {
console.log(event.sender, event.payload);
});
await client.postMessage("my-session", {
message: "Summarize this repository",
});
await session.prompt([{ type: "text", text: "Summarize this repository" }]);
off();
for await (const event of client.streamEvents("my-session")) {
console.log(event.type, event.data);
}
// Cleanup
await sandbox.stop();
```
## Authentication
Vercel Sandboxes support two authentication methods:
- **OIDC Token**: Set `VERCEL_OIDC_TOKEN` (recommended for CI/CD)
- **Access Token**: Set `VERCEL_ACCESS_TOKEN` (for local development, run `vercel env pull`)
See [Vercel Sandbox docs](https://vercel.com/docs/functions/sandbox) for details.
Vercel Sandboxes support OIDC token auth (recommended) and access-token auth.
See [Vercel Sandbox docs](https://vercel.com/docs/functions/sandbox).

23
docs/docs.json
View file

@ -50,8 +50,7 @@
"group": "Getting started",
"pages": [
"quickstart",
"building-chat-ui",
"manage-sessions",
"sdk-overview",
{
"group": "Deploy",
"icon": "server",
@ -68,11 +67,7 @@
]
},
{
"group": "SDKs",
"pages": ["sdks/typescript", "sdks/python"]
},
{
"group": "Agent Features",
"group": "Agent",
"pages": [
"agent-sessions",
"attachments",
@ -82,19 +77,24 @@
]
},
{
"group": "Features",
"group": "System",
"pages": ["file-system"]
},
{
"group": "Advanced",
"pages": ["advanced/acp-http-client"]
"group": "Orchestration",
"pages": [
"architecture",
"session-persistence",
"observability",
"multiplayer",
"security"
]
},
{
"group": "Reference",
"pages": [
"cli",
"inspector",
"session-transcript-schema",
"opencode-compatibility",
{
"group": "More",
@ -102,6 +102,7 @@
"credentials",
"daemon",
"cors",
"session-restoration",
"telemetry",
{
"group": "AI",

159
docs/file-system.mdx
View file

@ -5,183 +5,130 @@ sidebarTitle: "File System"
icon: "folder"
---
The filesystem API lets you list, read, write, move, and delete files inside the sandbox, plus upload batches of files via tar archives.
Control operations (`list`, `mkdir`, `move`, `stat`, `delete`) are ACP extensions on `/v2/rpc` and require an active ACP connection in the SDK.
The filesystem API lets you list, read, write, move, and delete files inside the sandbox, plus upload tar archives in batch.
Binary transfer is intentionally a separate HTTP API (not ACP extension methods):
- `GET /v2/fs/file`
- `PUT /v2/fs/file`
- `POST /v2/fs/upload-batch`
Reason: these are host/runtime capabilities implemented by Sandbox Agent for cross-agent-consistent behavior, and they may require streaming very large binary payloads that ACP JSON-RPC is not suited to transport efficiently.
This is intentionally separate from ACP native `fs/read_text_file` and `fs/write_text_file`.
ACP extension variants may exist in parallel for compatibility, but SDK defaults should use the HTTP endpoints above for binary transfer.
## Path Resolution
## Path resolution
- Absolute paths are used as-is.
- Relative paths use the session working directory when `sessionId` is provided.
- Without `sessionId`, relative paths resolve against the server home directory.
- Relative paths cannot contain `..` or absolute prefixes; requests that attempt to escape the root are rejected.
- Relative paths resolve from the server process working directory.
- Requests that attempt to escape allowed roots are rejected by the server.
The session working directory is the server process current working directory at the moment the session is created.
## List Entries
`listFsEntries()` uses ACP extension method `_sandboxagent/fs/list_entries`.
## List entries
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
const client = new SandboxAgentClient({ baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock" });
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
});
const entries = await client.listFsEntries({
const entries = await sdk.listFsEntries({
path: "./workspace",
sessionId: "my-session",
});
console.log(entries);
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v2/rpc" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "x-acp-connection-id: acp_conn_1" \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"_sandboxagent/fs/list_entries","params":{"path":"./workspace","sessionId":"my-session"}}'
curl -X GET "http://127.0.0.1:2468/v1/fs/entries?path=./workspace"
```
</CodeGroup>
## Read And Write Files
## Read and write files
`PUT /v2/fs/file` writes raw bytes. `GET /v2/fs/file` returns raw bytes.
`PUT /v1/fs/file` writes raw bytes. `GET /v1/fs/file` returns raw bytes.
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
const client = new SandboxAgentClient({ baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock" });
await client.writeFsFile({ path: "./notes.txt", sessionId: "my-session" }, "hello");
const bytes = await client.readFsFile({
path: "./notes.txt",
sessionId: "my-session",
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
});
await sdk.writeFsFile({ path: "./notes.txt" }, "hello");
const bytes = await sdk.readFsFile({ path: "./notes.txt" });
const text = new TextDecoder().decode(bytes);
console.log(text);
```
```bash cURL
curl -X PUT "http://127.0.0.1:2468/v2/fs/file?path=./notes.txt&sessionId=my-session" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
curl -X PUT "http://127.0.0.1:2468/v1/fs/file?path=./notes.txt" \
--data-binary "hello"
curl -X GET "http://127.0.0.1:2468/v2/fs/file?path=./notes.txt&sessionId=my-session" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
curl -X GET "http://127.0.0.1:2468/v1/fs/file?path=./notes.txt" \
--output ./notes.txt
```
</CodeGroup>
## Create Directories
`mkdirFs()` uses ACP extension method `_sandboxagent/fs/mkdir`.
## Create directories
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
const client = new SandboxAgentClient({ baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock" });
await client.mkdirFs({
path: "./data",
sessionId: "my-session",
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
});
await sdk.mkdirFs({ path: "./data" });
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v2/rpc" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "x-acp-connection-id: acp_conn_1" \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":2,"method":"_sandboxagent/fs/mkdir","params":{"path":"./data","sessionId":"my-session"}}'
curl -X POST "http://127.0.0.1:2468/v1/fs/mkdir?path=./data"
```
</CodeGroup>
## Move, Delete, And Stat
`moveFs()`, `statFs()`, and `deleteFsEntry()` use ACP extension methods (`_sandboxagent/fs/move`, `_sandboxagent/fs/stat`, `_sandboxagent/fs/delete_entry`).
## Move, delete, and stat
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
const client = new SandboxAgentClient({ baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock" });
await client.moveFs(
{ from: "./notes.txt", to: "./notes-old.txt", overwrite: true },
{ sessionId: "my-session" },
);
const stat = await client.statFs({
path: "./notes-old.txt",
sessionId: "my-session",
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
});
await client.deleteFsEntry({
path: "./notes-old.txt",
sessionId: "my-session",
await sdk.moveFs({
from: "./notes.txt",
to: "./notes-old.txt",
overwrite: true,
});
const stat = await sdk.statFs({ path: "./notes-old.txt" });
await sdk.deleteFsEntry({ path: "./notes-old.txt" });
console.log(stat);
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v2/rpc" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "x-acp-connection-id: acp_conn_1" \
curl -X POST "http://127.0.0.1:2468/v1/fs/move" \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":3,"method":"_sandboxagent/fs/move","params":{"from":"./notes.txt","to":"./notes-old.txt","overwrite":true,"sessionId":"my-session"}}'
-d '{"from":"./notes.txt","to":"./notes-old.txt","overwrite":true}'
curl -X POST "http://127.0.0.1:2468/v2/rpc" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "x-acp-connection-id: acp_conn_1" \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":4,"method":"_sandboxagent/fs/stat","params":{"path":"./notes-old.txt","sessionId":"my-session"}}'
curl -X GET "http://127.0.0.1:2468/v1/fs/stat?path=./notes-old.txt"
curl -X POST "http://127.0.0.1:2468/v2/rpc" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "x-acp-connection-id: acp_conn_1" \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":5,"method":"_sandboxagent/fs/delete_entry","params":{"path":"./notes-old.txt","sessionId":"my-session"}}'
curl -X DELETE "http://127.0.0.1:2468/v1/fs/entry?path=./notes-old.txt"
```
</CodeGroup>
## Batch Upload (Tar)
## Batch upload (tar)
Batch upload accepts `application/x-tar` only and extracts into the destination directory. The response returns absolute paths for extracted files, capped at 1024 entries.
Batch upload accepts `application/x-tar` and extracts into the destination directory.
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
import fs from "node:fs";
import path from "node:path";
import tar from "tar";
const client = new SandboxAgentClient({ baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock" });
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
});
const archivePath = path.join(process.cwd(), "skills.tar");
await tar.c({
@ -190,9 +137,8 @@ await tar.c({
}, ["."]);
const tarBuffer = await fs.promises.readFile(archivePath);
const result = await client.uploadFsBatch(tarBuffer, {
const result = await sdk.uploadFsBatch(tarBuffer, {
path: "./skills",
sessionId: "my-session",
});
console.log(result);
@ -201,8 +147,7 @@ console.log(result);
```bash cURL
tar -cf skills.tar -C ./skills .
curl -X POST "http://127.0.0.1:2468/v2/fs/upload-batch?path=./skills&sessionId=my-session" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
curl -X POST "http://127.0.0.1:2468/v1/fs/upload-batch?path=./skills" \
-H "Content-Type: application/x-tar" \
--data-binary @skills.tar
```

265
docs/manage-sessions.mdx
View file

@ -1,265 +0,0 @@
---
title: "Manage Sessions"
description: "Persist and replay agent transcripts across connections."
icon: "database"
---
Sandbox Agent stores sessions in memory only. When the server restarts or the sandbox is destroyed, all session data is lost. It's your responsibility to persist events to your own database.
See the [Building a Chat UI](/building-chat-ui) guide for understanding session lifecycle events like `session.started` and `session.ended`.
## Recommended approach
1. Store events to your database as they arrive
2. On reconnect, get the last event's `sequence` and pass it as `offset`
3. The API returns events where `sequence > offset`
This prevents duplicate writes and lets you recover from disconnects.
## Receiving Events
Two ways to receive events: SSE streaming (recommended) or polling.
### Streaming
Use SSE for real-time events with automatic reconnection support.
```typescript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
agent: "mock",
});
// Get offset from last stored event (0 returns all events)
const lastEvent = await db.getLastEvent("my-session");
const offset = lastEvent?.sequence ?? 0;
// Stream from where you left off
for await (const event of client.streamEvents("my-session", { offset })) {
await db.insertEvent("my-session", event);
}
```
### Polling
If you can't use SSE streaming, poll the events endpoint:
```typescript
const lastEvent = await db.getLastEvent("my-session");
let offset = lastEvent?.sequence ?? 0;
while (true) {
const { events } = await client.getEvents("my-session", {
offset,
limit: 100
});
for (const event of events) {
await db.insertEvent("my-session", event);
offset = event.sequence;
}
await sleep(1000);
}
```
## Database options
Choose where to persist events based on your requirements. For most use cases, we recommend Rivet Actors.
| | Durable | Real-time | Multiplayer | Scaling | Throughput | Complexity |
|---------|:-------:|:---------:|:-----------:|---------|------------|------------|
| Rivet Actors | ✓ | ✓ | ✓ | Auto-sharded, one actor per session | Millions of concurrent sessions | Zero infrastructure |
| PostgreSQL | ✓ | | | Manual sharding | Connection pool limited | Connection pools, migrations |
| Redis | | ✓ | | Redis Cluster | High, in-memory | Memory management, Sentinel for failover |
### Rivet Actors
For production workloads, [Rivet Actors](https://rivet.gg) provide a managed solution for:
- **Persistent state**: Events survive crashes and restarts
- **Real-time streaming**: Built-in WebSocket support for clients
- **Horizontal scaling**: Run thousands of concurrent sessions
- **Observability**: Built-in logging and metrics
#### Actor
```typescript
import { actor } from "rivetkit";
import { Daytona } from "@daytonaio/sdk";
import { SandboxAgent, SandboxAgentClient, AgentEvent } from "sandbox-agent";
interface CodingSessionState {
sandboxId: string;
baseUrl: string;
sessionId: string;
events: AgentEvent[];
}
interface CodingSessionVars {
client: SandboxAgentClient;
}
const daytona = new Daytona();
const codingSession = actor({
createState: async (): Promise<CodingSessionState> => {
const sandbox = await daytona.create({
snapshot: "sandbox-agent-ready",
envVars: {
ANTHROPIC_API_KEY: process.env.ANTHROPIC_API_KEY,
OPENAI_API_KEY: process.env.OPENAI_API_KEY,
},
autoStopInterval: 0,
});
await sandbox.process.executeCommand(
"nohup sandbox-agent server --no-token --host 0.0.0.0 --port 3000 &"
);
const baseUrl = (await sandbox.getSignedPreviewUrl(3000)).url;
const sessionId = crypto.randomUUID();
return {
sandboxId: sandbox.id,
baseUrl,
sessionId,
events: [],
};
},
createVars: async (c): Promise<CodingSessionVars> => {
const client = new SandboxAgentClient({
baseUrl: c.state.baseUrl,
agent: "mock",
});
await client.createSession(c.state.sessionId, { agent: "claude" });
return { client };
},
onDestroy: async (c) => {
const sandbox = await daytona.get(c.state.sandboxId);
await sandbox.delete();
},
run: async (c) => {
for await (const event of c.vars.client.streamEvents(c.state.sessionId)) {
c.state.events.push(event);
c.broadcast("agentEvent", event);
}
},
actions: {
postMessage: async (c, message: string) => {
await c.vars.client.postMessage(c.state.sessionId, message);
},
getTranscript: (c) => c.state.events,
},
});
```
#### Client
<CodeGroup>
```typescript TypeScript
import { createClient } from "rivetkit/client";
const client = createClient();
const session = client.codingSession.getOrCreate(["my-session"]);
const conn = session.connect();
conn.on("agentEvent", (event) => {
console.log(event.type, event.data);
});
await conn.postMessage("Create a new React component for user profiles");
const transcript = await conn.getTranscript();
```
```typescript React
import { createRivetKit } from "@rivetkit/react";
const { useActor } = createRivetKit();
function CodingSession() {
const [messages, setMessages] = useState<AgentEvent[]>([]);
const session = useActor({ name: "codingSession", key: ["my-session"] });
session.useEvent("agentEvent", (event) => {
setMessages((prev) => [...prev, event]);
});
const sendPrompt = async (prompt: string) => {
await session.connection?.postMessage(prompt);
};
return (
<div>
{messages.map((msg, i) => (
<div key={i}>{JSON.stringify(msg)}</div>
))}
<button onClick={() => sendPrompt("Build a login page")}>
Send Prompt
</button>
</div>
);
}
```
</CodeGroup>
### PostgreSQL
```sql
CREATE TABLE agent_events (
event_id TEXT PRIMARY KEY,
session_id TEXT NOT NULL,
native_session_id TEXT,
sequence INTEGER NOT NULL,
time TIMESTAMPTZ NOT NULL,
type TEXT NOT NULL,
source TEXT NOT NULL,
synthetic BOOLEAN NOT NULL DEFAULT FALSE,
data JSONB NOT NULL,
UNIQUE(session_id, sequence)
);
CREATE INDEX idx_events_session ON agent_events(session_id, sequence);
```
### Redis
```typescript
// Append event to list
await redis.rpush(`session:${sessionId}`, JSON.stringify(event));
// Get events from offset
const events = await redis.lrange(`session:${sessionId}`, offset, -1);
```
## Handling disconnects
The SSE stream may disconnect due to network issues. Handle reconnection gracefully:
```typescript
async function streamWithRetry(sessionId: string) {
while (true) {
try {
const lastEvent = await db.getLastEvent(sessionId);
const offset = lastEvent?.sequence ?? 0;
for await (const event of client.streamEvents(sessionId, { offset })) {
await db.insertEvent(sessionId, event);
}
} catch (error) {
console.error("Stream disconnected, reconnecting...", error);
await sleep(1000);
}
}
}
```

141
docs/mcp-config.mdx
View file

@ -5,119 +5,80 @@ sidebarTitle: "MCP"
icon: "plug"
---
MCP (Model Context Protocol) servers extend agents with tools. Sandbox Agent can auto-load MCP servers when a session starts by passing an `mcp` map in the create-session request.
MCP (Model Context Protocol) servers extend agents with tools and external context.
## Session Config
## Configuring MCP servers
The `mcp` field is a map of server name to config. Use `type: "local"` for stdio servers and `type: "remote"` for HTTP/SSE servers:
The HTTP config endpoints let you store/retrieve MCP server configs by directory + name.
<CodeGroup>
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
await client.createSession("claude-mcp", {
agent: "claude",
mcp: {
filesystem: {
type: "local",
command: "my-mcp-server",
args: ["--root", "."],
```ts
// Create MCP config
await sdk.setMcpConfig(
{
directory: "/workspace",
mcpName: "github",
},
github: {
{
type: "remote",
url: "https://example.com/mcp",
headers: {
Authorization: "Bearer ${GITHUB_TOKEN}",
},
},
);
// Create a session using the configured MCP servers
const session = await sdk.createSession({
agent: "claude",
sessionInit: {
cwd: "/workspace",
},
});
await session.prompt([
{ type: "text", text: "Use available MCP servers to help with this task." },
]);
// List MCP configs
const config = await sdk.getMcpConfig({
directory: "/workspace",
mcpName: "github",
});
console.log(config.type);
// Delete MCP config
await sdk.deleteMcpConfig({
directory: "/workspace",
mcpName: "github",
});
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/claude-mcp" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"agent": "claude",
"mcp": {
"filesystem": {
"type": "local",
"command": "my-mcp-server",
"args": ["--root", "."]
},
"github": {
"type": "remote",
"url": "https://example.com/mcp",
"headers": {
"Authorization": "Bearer ${GITHUB_TOKEN}"
}
}
}
}'
```
## Config fields
</CodeGroup>
## Config Fields
### Local Server
Stdio servers that run inside the sandbox.
### Local server
| Field | Description |
|---|---|
| `type` | `local` |
| `command` | string or array (`["node", "server.js"]`) |
| `args` | array of string arguments |
| `env` | environment variables map |
| `enabled` | enable or disable the server |
| `timeoutMs` | tool timeout override |
| `cwd` | working directory for the MCP process |
| `command` | executable path |
| `args` | array of CLI args |
| `env` | environment variable map |
| `cwd` | working directory |
| `enabled` | enable/disable server |
| `timeoutMs` | timeout override |
```json
{
"type": "local",
"command": ["node", "./mcp/server.js"],
"args": ["--root", "."],
"env": { "LOG_LEVEL": "debug" },
"cwd": "/workspace"
}
```
### Remote Server
HTTP/SSE servers accessed over the network.
### Remote server
| Field | Description |
|---|---|
| `type` | `remote` |
| `url` | MCP server URL |
| `headers` | static headers map |
| `bearerTokenEnvVar` | env var name to inject into `Authorization: Bearer ...` |
| `envHeaders` | map of header name to env var name |
| `oauth` | object with `clientId`, `clientSecret`, `scope`, or `false` to disable |
| `enabled` | enable or disable the server |
| `timeoutMs` | tool timeout override |
| `transport` | `http` or `sse` |
| `headers` | static headers map |
| `bearerTokenEnvVar` | env var name to inject in auth header |
| `envHeaders` | header name to env var map |
| `oauth` | optional OAuth config object |
| `enabled` | enable/disable server |
| `timeoutMs` | timeout override |
```json
{
"type": "remote",
"url": "https://example.com/mcp",
"headers": { "x-client": "sandbox-agent" },
"bearerTokenEnvVar": "MCP_TOKEN",
"transport": "sse"
}
```
## Custom MCP Servers
## Custom MCP servers
To bundle and upload your own MCP server into the sandbox, see [Custom Tools](/custom-tools).

115
docs/multiplayer.mdx Normal file
View file

@ -0,0 +1,115 @@
---
title: "Multiplayer"
description: "Use Rivet Actors to coordinate shared sessions."
icon: "users"
---
For multiplayer orchestration, use [Rivet Actors](https://rivet.dev/docs/actors).
Recommended model:
- One actor per collaborative workspace/thread.
- The actor owns Sandbox Agent session lifecycle and persistence.
- Clients connect to the actor and receive realtime broadcasts.
Use [actor keys](https://rivet.dev/docs/actors/keys) to map each workspace to one actor, [events](https://rivet.dev/docs/actors/events) for realtime updates, and [lifecycle hooks](https://rivet.dev/docs/actors/lifecycle) for cleanup.
## Example
<CodeGroup>
```ts Actor (server)
import { actor, setup } from "rivetkit";
import { SandboxAgent } from "sandbox-agent";
import { RivetSessionPersistDriver, type RivetPersistState } from "@sandbox-agent/persist-rivet";
type WorkspaceState = RivetPersistState & {
sandboxId: string;
baseUrl: string;
};
export const workspace = actor({
createState: async () => {
return {
sandboxId: "sbx_123",
baseUrl: "http://127.0.0.1:2468",
} satisfies Partial<WorkspaceState>;
},
createVars: async (c) => {
const persist = new RivetSessionPersistDriver(c);
const sdk = await SandboxAgent.connect({
baseUrl: c.state.baseUrl,
persist,
});
const session = await sdk.resumeOrCreateSession({ id: "default", agent: "codex" });
const unsubscribe = session.onEvent((event) => {
c.broadcast("session.event", event);
});
return { sdk, session, unsubscribe };
},
actions: {
getSessionInfo: (c) => ({
workspaceId: c.key[0],
sandboxId: c.state.sandboxId,
}),
prompt: async (c, input: { userId: string; text: string }) => {
c.broadcast("chat.user", {
userId: input.userId,
text: input.text,
createdAt: Date.now(),
});
await c.vars.session.prompt([{ type: "text", text: input.text }]);
},
},
onSleep: async (c) => {
c.vars.unsubscribe?.();
await c.vars.sdk.dispose();
},
});
export const registry = setup({
use: { workspace },
});
```
```ts Client (browser)
import { createClient } from "rivetkit/client";
import type { registry } from "./actors";
const client = createClient<typeof registry>({
endpoint: process.env.NEXT_PUBLIC_RIVET_ENDPOINT!,
});
const workspaceId = "workspace-42";
const room = client.workspace.getOrCreate([workspaceId]);
const conn = room.connect();
conn.on("chat.user", (event) => {
console.log("user message", event);
});
conn.on("session.event", (event) => {
console.log("sandbox event", event);
});
await conn.prompt({
userId: "user-123",
text: "Propose a refactor plan for auth middleware.",
});
```
</CodeGroup>
## Notes
- Keep sandbox calls actor-only. Browser clients should not call Sandbox Agent directly.
- Use `@sandbox-agent/persist-rivet` so session history persists in actor state.
- For client connection patterns, see [Rivet JavaScript client](https://rivet.dev/docs/clients/javascript).

64
docs/observability.mdx Normal file
View file

@ -0,0 +1,64 @@
---
title: "Observability"
description: "Track session activity with OpenTelemetry."
icon: "terminal"
---
Use OpenTelemetry to instrument session traffic, then ship telemetry to your collector/backend.
## Common collectors and backends
- [OpenTelemetry Collector](https://opentelemetry.io/docs/collector/)
- [Jaeger](https://www.jaegertracing.io/)
- [Grafana Tempo](https://grafana.com/oss/tempo/)
- [Honeycomb](https://www.honeycomb.io/)
- [Datadog APM](https://docs.datadoghq.com/tracing/)
## Example: trace a prompt round-trip
Wrap `session.prompt()` in a span to measure the full round-trip, then log individual events as span events.
Assumes your OTEL provider/exporter is already configured.
```ts
import { trace } from "@opentelemetry/api";
import { SandboxAgent } from "sandbox-agent";
const tracer = trace.getTracer("my-app/sandbox-agent");
const sdk = await SandboxAgent.connect({
baseUrl: process.env.SANDBOX_URL!,
});
const session = await sdk.createSession({ agent: "mock" });
// Log each event as an OTEL span event on the active span
const unsubscribe = session.onEvent((event) => {
const activeSpan = trace.getActiveSpan();
if (!activeSpan) return;
activeSpan.addEvent("session.event", {
"sandbox.sender": event.sender,
"sandbox.event_index": event.eventIndex,
});
});
// The span covers the full prompt round-trip
await tracer.startActiveSpan("sandbox_agent.prompt", async (span) => {
span.setAttribute("sandbox.session_id", session.id);
try {
const result = await session.prompt([
{ type: "text", text: "Summarize this repository." },
]);
span.setAttribute("sandbox.stop_reason", result.stopReason);
} catch (error) {
span.recordException(error as Error);
throw error;
} finally {
span.end();
}
});
unsubscribe();
```

1413
docs/openapi.json
View file

File diff suppressed because it is too large Load diff

129
docs/opencode-compatibility.mdx
View file

@ -1,26 +1,125 @@
---
title: "OpenCode Compatibility"
description: "Status of the OpenCode bridge during ACP v2 migration."
description: "Connect OpenCode clients, SDKs, and web UI to Sandbox Agent."
---
OpenCode compatibility is intentionally deferred during ACP core migration.
<Warning>
**Experimental**: OpenCode SDK/UI compatibility may change.
</Warning>
## Current status (v2 core phases)
Sandbox Agent exposes an OpenCode-compatible API at `/opencode`.
- `/opencode/*` routes are disabled.
- `sandbox-agent opencode` returns an explicit disabled error.
- This is expected while ACP runtime, SDK, and inspector migration is completed.
## Why use OpenCode clients with Sandbox Agent?
## Planned re-enable step
- OpenCode CLI (`opencode attach`)
- OpenCode web UI
- OpenCode TypeScript SDK (`@opencode-ai/sdk`)
OpenCode support is restored in a dedicated phase after ACP core is stable:
## Quick start
1. Reintroduce `/opencode/*` routing on top of ACP internals.
2. Add dedicated OpenCode ↔ ACP integration tests.
3. Re-enable OpenCode docs and operational guidance.
### OpenCode CLI / TUI
Track details in:
```bash
sandbox-agent opencode --port 2468 --no-token
```
- `research/acp/spec.md`
- `research/acp/migration-steps.md`
- `research/acp/todo.md`
Or start server + attach manually:
```bash
sandbox-agent server --no-token --host 127.0.0.1 --port 2468
opencode attach http://localhost:2468/opencode
```
With authentication enabled:
```bash
sandbox-agent server --token "$SANDBOX_TOKEN" --host 127.0.0.1 --port 2468
opencode attach http://localhost:2468/opencode --password "$SANDBOX_TOKEN"
```
### OpenCode web UI
<Steps>
<Step title="Start Sandbox Agent with CORS">
```bash
sandbox-agent server --no-token --host 127.0.0.1 --port 2468 --cors-allow-origin http://127.0.0.1:5173
```
</Step>
<Step title="Run OpenCode web app">
```bash
git clone https://github.com/anomalyco/opencode
cd opencode/packages/app
export VITE_OPENCODE_SERVER_HOST=127.0.0.1
export VITE_OPENCODE_SERVER_PORT=2468
bun install
bun run dev -- --host 127.0.0.1 --port 5173
```
</Step>
<Step title="Open UI">
Visit `http://127.0.0.1:5173/`.
</Step>
</Steps>
### OpenCode SDK
```typescript
import { createOpencodeClient } from "@opencode-ai/sdk";
const client = createOpencodeClient({
baseUrl: "http://localhost:2468/opencode",
});
const session = await client.session.create();
await client.session.promptAsync({
path: { id: session.data.id },
body: {
parts: [{ type: "text", text: "Hello, write a hello world script" }],
},
});
const events = await client.event.subscribe({});
for await (const event of events.stream) {
console.log(event);
}
```
## Notes
- API base path: `/opencode`
- If server auth is enabled, pass bearer auth (or `--password` in OpenCode CLI)
- For browser UIs, configure CORS with `--cors-allow-origin`
- Provider selector currently exposes compatible providers (`mock`, `amp`, `claude`, `codex`)
- Provider/model metadata for compatibility endpoints is normalized and may differ from native OpenCode grouping
- Optional proxy: set `OPENCODE_COMPAT_PROXY_URL` to forward selected endpoints to native OpenCode
## Endpoint coverage
<Accordion title="Endpoint Status Table">
| Endpoint | Status | Notes |
|---|---|---|
| `GET /event` | ✓ | Session/message updates (SSE) |
| `GET /global/event` | ✓ | GlobalEvent-wrapped stream |
| `GET /session` | ✓ | Session list |
| `POST /session` | ✓ | Create session |
| `GET /session/{id}` | ✓ | Session details |
| `POST /session/{id}/message` | ✓ | Send message |
| `GET /session/{id}/message` | ✓ | Session messages |
| `GET /permission` | ✓ | Pending permissions |
| `POST /permission/{id}/reply` | ✓ | Permission reply |
| `GET /question` | ✓ | Pending questions |
| `POST /question/{id}/reply` | ✓ | Question reply |
| `GET /provider` | ✓ | Provider metadata |
| `GET /command` | ↔ | Proxied when `OPENCODE_COMPAT_PROXY_URL` is set; otherwise stub |
| `GET /config` | ↔ | Proxied when set; otherwise stub |
| `PATCH /config` | ↔ | Proxied when set; otherwise local compatibility behavior |
| `GET /global/config` | ↔ | Proxied when set; otherwise stub |
| `PATCH /global/config` | ↔ | Proxied when set; otherwise local compatibility behavior |
| `/tui/*` | ↔ | Proxied when set; otherwise local compatibility behavior |
| `GET /agent` | − | Agent list |
| *other endpoints* | − | Empty/stub responses |
✓ Functional ↔ Proxied optional − Stubbed
</Accordion>

174
docs/quickstart.mdx
View file

@ -70,10 +70,10 @@ icon: "rocket"
<AccordionGroup>
<Accordion title="Extracting API keys from current machine">
Use `sandbox-agent credentials extract-env --export` to extract your existing API keys (Anthropic, OpenAI, etc.) from your existing Claude Code or Codex config files on your machine.
Use `sandbox-agent credentials extract-env --export` to extract your existing API keys (Anthropic, OpenAI, etc.) from local Claude Code or Codex config files.
</Accordion>
<Accordion title="Testing without API keys">
If you want to test Sandbox Agent without API keys, use the `mock` agent to test the SDK without any credentials. It simulates agent responses for development and testing.
Use the `mock` agent for SDK and integration testing without provider credentials.
</Accordion>
</AccordionGroup>
</Step>
@ -84,7 +84,7 @@ icon: "rocket"
Install and run the binary directly.
```bash
curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh
curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh
sandbox-agent server --no-token --host 0.0.0.0 --port 2468
```
</Tab>
@ -93,7 +93,7 @@ icon: "rocket"
Run without installing globally.
```bash
npx @sandbox-agent/cli server --no-token --host 0.0.0.0 --port 2468
npx @sandbox-agent/cli@0.2.x server --no-token --host 0.0.0.0 --port 2468
```
</Tab>
@ -101,7 +101,7 @@ icon: "rocket"
Run without installing globally.
```bash
bunx @sandbox-agent/cli server --no-token --host 0.0.0.0 --port 2468
bunx @sandbox-agent/cli@0.2.x server --no-token --host 0.0.0.0 --port 2468
```
</Tab>
@ -109,7 +109,7 @@ icon: "rocket"
Install globally, then run.
```bash
npm install -g @sandbox-agent/cli
npm install -g @sandbox-agent/cli@0.2.x
sandbox-agent server --no-token --host 0.0.0.0 --port 2468
```
</Tab>
@ -118,33 +118,32 @@ icon: "rocket"
Install globally, then run.
```bash
bun add -g @sandbox-agent/cli
bun add -g @sandbox-agent/cli@0.2.x
# Allow Bun to run postinstall scripts for native binaries (required for SandboxAgent.start()).
bun pm -g trust @sandbox-agent/cli-linux-x64 @sandbox-agent/cli-darwin-arm64 @sandbox-agent/cli-darwin-x64 @sandbox-agent/cli-win32-x64
sandbox-agent server --no-token --host 0.0.0.0 --port 2468
```
</Tab>
<Tab title="Node.js (local)">
For local development, use `SandboxAgent.start()` to automatically spawn and manage the server as a subprocess.
For local development, use `SandboxAgent.start()` to spawn and manage the server as a subprocess.
```bash
npm install sandbox-agent
npm install sandbox-agent@0.2.x
```
```typescript
import { SandboxAgent } from "sandbox-agent";
const client = await SandboxAgent.start();
const sdk = await SandboxAgent.start();
```
</Tab>
<Tab title="Bun (local)">
For local development, use `SandboxAgent.start()` to automatically spawn and manage the server as a subprocess.
For local development, use `SandboxAgent.start()` to spawn and manage the server as a subprocess.
```bash
bun add sandbox-agent
bun add sandbox-agent@0.2.x
# Allow Bun to run postinstall scripts for native binaries (required for SandboxAgent.start()).
bun pm trust @sandbox-agent/cli-linux-x64 @sandbox-agent/cli-darwin-arm64 @sandbox-agent/cli-darwin-x64 @sandbox-agent/cli-win32-x64
```
@ -152,10 +151,8 @@ icon: "rocket"
```typescript
import { SandboxAgent } from "sandbox-agent";
const client = await SandboxAgent.start();
const sdk = await SandboxAgent.start();
```
This installs the binary and starts the server for you. No manual setup required.
</Tab>
<Tab title="Build from source">
@ -171,9 +168,9 @@ icon: "rocket"
<AccordionGroup>
<Accordion title="Configuring token">
Tokens are usually not required. Most sandbox providers (E2B, Daytona, etc.) already secure their networking at the infrastructure level, so the server endpoint is never publicly accessible. For local development, binding to `127.0.0.1` ensures only local connections are accepted.
Tokens are usually not required. Most sandbox providers (E2B, Daytona, etc.) already secure networking at the infrastructure layer.
If you need to expose the server on a public endpoint, use `--token "$SANDBOX_TOKEN"` to require authentication on all requests:
If you expose the server publicly, use `--token "$SANDBOX_TOKEN"` to require authentication:
```bash
sandbox-agent server --token "$SANDBOX_TOKEN" --host 0.0.0.0 --port 2468
@ -184,34 +181,32 @@ icon: "rocket"
<Tabs>
<Tab title="TypeScript">
```typescript
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
const client = new SandboxAgentClient({
const sdk = await SandboxAgent.connect({
baseUrl: "http://your-server:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
```
</Tab>
<Tab title="curl">
```bash
curl "http://your-server:2468/v1/sessions" \
curl "http://your-server:2468/v1/health" \
-H "Authorization: Bearer $SANDBOX_TOKEN"
```
</Tab>
<Tab title="CLI">
```bash
sandbox-agent api sessions list \
--endpoint http://your-server:2468 \
--token "$SANDBOX_TOKEN"
sandbox-agent --token "$SANDBOX_TOKEN" api agents list \
--endpoint http://your-server:2468
```
</Tab>
</Tabs>
</Accordion>
<Accordion title="CORS">
If you're calling the server from a browser, see the [CORS configuration guide](/docs/cors).
If you're calling the server from a browser, see the [CORS configuration guide](/cors).
</Accordion>
</AccordionGroup>
</Step>
@ -226,124 +221,57 @@ icon: "rocket"
sandbox-agent install-agent amp
```
If agents are not installed up front, they will be lazily installed when creating a session. It's recommended to pre-install agents then take a snapshot of the sandbox for faster coldstarts.
If agents are not installed up front, they are lazily installed when creating a session.
</Step>
<Step title="Create a session">
<Tabs>
<Tab title="TypeScript">
```typescript
import { SandboxAgentClient } from "sandbox-agent";
import { SandboxAgent } from "sandbox-agent";
const client = new SandboxAgentClient({
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
agent: "claude",
});
await client.createSession("my-session", {
const session = await sdk.createSession({
agent: "claude",
agentMode: "build",
permissionMode: "default",
sessionInit: {
cwd: "/",
mcpServers: [],
},
});
```
</Tab>
<Tab title="curl">
```bash
curl -X POST "http://127.0.0.1:2468/v1/sessions/my-session" \
-H "Content-Type: application/json" \
-d '{"agent":"claude","agentMode":"build","permissionMode":"default"}'
console.log(session.id);
```
</Tab>
<Tab title="CLI">
```bash
sandbox-agent api sessions create my-session \
--agent claude \
--endpoint http://127.0.0.1:2468
```
</Tab>
</Tabs>
</Step>
<Step title="Send a message">
<Tabs>
<Tab title="TypeScript">
```typescript
await client.postMessage("my-session", {
message: "Summarize the repository and suggest next steps.",
});
```
</Tab>
const result = await session.prompt([
{ type: "text", text: "Summarize the repository and suggest next steps." },
]);
<Tab title="curl">
```bash
curl -X POST "http://127.0.0.1:2468/v1/sessions/my-session/messages" \
-H "Content-Type: application/json" \
-d '{"message":"Summarize the repository and suggest next steps."}'
console.log(result.stopReason);
```
</Tab>
<Tab title="CLI">
```bash
sandbox-agent api sessions send-message my-session \
--message "Summarize the repository and suggest next steps." \
--endpoint http://127.0.0.1:2468
```
</Tab>
</Tabs>
</Step>
<Step title="Read events">
<Tabs>
<Tab title="TypeScript">
```typescript
// Poll for events
const events = await client.getEvents("my-session", { offset: 0, limit: 50 });
const off = session.onEvent((event) => {
console.log(event.sender, event.payload);
});
// Or stream events
for await (const event of client.streamEvents("my-session", { offset: 0 })) {
console.log(event.type, event.data);
}
const page = await sdk.getEvents({
sessionId: session.id,
limit: 50,
});
console.log(page.items.length);
off();
```
</Tab>
<Tab title="curl">
```bash
# Poll for events
curl "http://127.0.0.1:2468/v1/sessions/my-session/events?offset=0&limit=50"
# Stream events via SSE
curl "http://127.0.0.1:2468/v1/sessions/my-session/events/sse?offset=0"
# Single-turn stream (post message and get streamed response)
curl -N -X POST "http://127.0.0.1:2468/v1/sessions/my-session/messages/stream" \
-H "Content-Type: application/json" \
-d '{"message":"Hello"}'
```
</Tab>
<Tab title="CLI">
```bash
# Poll for events
sandbox-agent api sessions events my-session \
--endpoint http://127.0.0.1:2468
# Stream events via SSE
sandbox-agent api sessions events-sse my-session \
--endpoint http://127.0.0.1:2468
# Single-turn stream
sandbox-agent api sessions send-message-stream my-session \
--message "Hello" \
--endpoint http://127.0.0.1:2468
```
</Tab>
</Tabs>
</Step>
<Step title="Test with Inspector">
Open the Inspector UI at `/ui/` on your server (e.g., `http://localhost:2468/ui/`) to inspect session state using a GUI.
Open the Inspector UI at `/ui/` on your server (for example, `http://localhost:2468/ui/`) to inspect sessions and events in a GUI.
<Frame>
<img src="/images/inspector.png" alt="Sandbox Agent Inspector" />
@ -354,13 +282,13 @@ icon: "rocket"
## Next steps
<CardGroup cols={3}>
<Card title="Build a Chat UI" icon="comments" href="/building-chat-ui">
Learn how to build a chat interface for your agent.
<Card title="Session Persistence" icon="database" href="/session-persistence">
Configure in-memory, Rivet Actor state, IndexedDB, SQLite, and Postgres persistence.
</Card>
<Card title="Manage Sessions" icon="database" href="/manage-sessions">
Persist and replay agent transcripts.
<Card title="Deploy to a Sandbox" icon="box" href="/deploy/local">
Deploy your agent to E2B, Daytona, Docker, Vercel, or Cloudflare.
</Card>
<Card title="Deploy to a Sandbox" icon="box" href="/deploy">
Deploy your agent to E2B, Daytona, or Vercel Sandboxes.
<Card title="SDK Overview" icon="compass" href="/sdk-overview">
Use the latest TypeScript SDK API.
</Card>
</CardGroup>

174
docs/sdk-overview.mdx Normal file
View file

@ -0,0 +1,174 @@
---
title: "SDK Overview"
description: "Use the TypeScript SDK to manage Sandbox Agent sessions and APIs."
icon: "compass"
---
The TypeScript SDK is centered on `sandbox-agent` and its `SandboxAgent` class.
## Install
<Tabs>
<Tab title="npm">
```bash
npm install sandbox-agent@0.2.x
```
</Tab>
<Tab title="bun">
```bash
bun add sandbox-agent@0.2.x
# Allow Bun to run postinstall scripts for native binaries (required for SandboxAgent.start()).
bun pm trust @sandbox-agent/cli-linux-x64 @sandbox-agent/cli-darwin-arm64 @sandbox-agent/cli-darwin-x64 @sandbox-agent/cli-win32-x64
```
</Tab>
</Tabs>
## Optional persistence drivers
```bash
npm install @sandbox-agent/persist-indexeddb@0.2.x @sandbox-agent/persist-sqlite@0.2.x @sandbox-agent/persist-postgres@0.2.x
```
## Create a client
```ts
import { SandboxAgent } from "sandbox-agent";
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
});
```
With persistence:
```ts
import { SandboxAgent } from "sandbox-agent";
import { SQLiteSessionPersistDriver } from "@sandbox-agent/persist-sqlite";
const persist = new SQLiteSessionPersistDriver({
filename: "./sessions.db",
});
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
persist,
});
```
Local autospawn (Node.js only):
```ts
import { SandboxAgent } from "sandbox-agent";
const localSdk = await SandboxAgent.start();
await localSdk.dispose();
```
## Session flow
```ts
const session = await sdk.createSession({
agent: "mock",
sessionInit: {
cwd: "/",
mcpServers: [],
},
});
const prompt = await session.prompt([
{ type: "text", text: "Summarize this repository." },
]);
console.log(prompt.stopReason);
```
Load and destroy:
```ts
const restored = await sdk.resumeSession(session.id);
await restored.prompt([{ type: "text", text: "Continue from previous context." }]);
await sdk.destroySession(restored.id);
```
## Events
Subscribe to live events:
```ts
const unsubscribe = session.onEvent((event) => {
console.log(event.eventIndex, event.sender, event.payload);
});
await session.prompt([{ type: "text", text: "Give me a short summary." }]);
unsubscribe();
```
Fetch persisted events:
```ts
const page = await sdk.getEvents({
sessionId: session.id,
limit: 100,
});
console.log(page.items.length);
```
## Control-plane and HTTP helpers
```ts
const health = await sdk.getHealth();
const agents = await sdk.listAgents();
await sdk.installAgent("codex", { reinstall: true });
const entries = await sdk.listFsEntries({ path: "." });
const writeResult = await sdk.writeFsFile({ path: "./hello.txt" }, "hello");
console.log(health.status, agents.agents.length, entries.length, writeResult.path);
```
## Error handling
```ts
import { SandboxAgentError } from "sandbox-agent";
try {
await sdk.listAgents();
} catch (error) {
if (error instanceof SandboxAgentError) {
console.error(error.status, error.problem);
}
}
```
## Inspector URL
```ts
import { buildInspectorUrl } from "sandbox-agent";
const url = buildInspectorUrl({
baseUrl: "https://your-sandbox-agent.example.com",
headers: { "X-Custom-Header": "value" },
});
console.log(url);
```
Parameters:
- `baseUrl` (required): Sandbox Agent server URL
- `token` (optional): Bearer token for authenticated servers
- `headers` (optional): Additional request headers
## Types
```ts
import type {
AgentInfo,
HealthResponse,
SessionEvent,
SessionRecord,
} from "sandbox-agent";
```

41
docs/sdks/python.mdx
View file

@ -1,41 +0,0 @@
---
title: "Python"
description: "Python client for managing sessions and streaming events."
icon: "python"
tag: "Coming Soon"
---
The Python SDK is on our roadmap. It will provide a typed client for managing sessions and streaming events, similar to the TypeScript SDK.
In the meantime, you can use the [HTTP API](/http-api) directly with any HTTP client like `requests` or `httpx`.
```python
import httpx
base_url = "http://127.0.0.1:2468"
headers = {"Authorization": f"Bearer {token}"}
# Create a session
httpx.post(
f"{base_url}/v1/sessions/my-session",
headers=headers,
json={"agent": "claude", "permissionMode": "default"}
)
# Send a message
httpx.post(
f"{base_url}/v1/sessions/my-session/messages",
headers=headers,
json={"message": "Hello from Python"}
)
# Get events
response = httpx.get(
f"{base_url}/v1/sessions/my-session/events",
headers=headers,
params={"offset": 0, "limit": 50}
)
events = response.json()["events"]
```
Want the Python SDK sooner? [Open an issue](https://github.com/rivet-dev/sandbox-agent/issues) to let us know.

267
docs/sdks/typescript.mdx
View file

@ -1,267 +0,0 @@
---
title: "TypeScript"
description: "Use the TypeScript SDK to manage ACP sessions and Sandbox Agent HTTP APIs."
icon: "js"
---
The TypeScript SDK is centered on `sandbox-agent` and its `SandboxAgentClient`, which provides a Sandbox-facing API for session flows, ACP extensions, and binary HTTP filesystem helpers.
## Install
<Tabs>
<Tab title="npm">
```bash
npm install sandbox-agent
```
</Tab>
<Tab title="bun">
```bash
bun add sandbox-agent
# Allow Bun to run postinstall scripts for native binaries (required for SandboxAgent.start()).
bun pm trust @sandbox-agent/cli-linux-x64 @sandbox-agent/cli-darwin-arm64 @sandbox-agent/cli-darwin-x64 @sandbox-agent/cli-win32-x64
```
</Tab>
</Tabs>
## Create a client
```ts
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
```
`SandboxAgentClient` is the canonical API. By default it auto-connects (`autoConnect: true`), so provide `agent` in the constructor. Use the instance method `client.connect()` only when you explicitly set `autoConnect: false`.
## Autospawn (Node only)
If you run locally, the SDK can launch the server for you.
```ts
import { SandboxAgent } from "sandbox-agent";
const client = await SandboxAgent.start({
agent: "mock",
});
await client.dispose();
```
Autospawn uses the local `sandbox-agent` binary. Install `@sandbox-agent/cli` (recommended) or set
`SANDBOX_AGENT_BIN` to a custom path.
## Connect lifecycle
Use manual mode when you want explicit ACP session lifecycle control.
```ts
import {
AlreadyConnectedError,
NotConnectedError,
SandboxAgentClient,
} from "sandbox-agent";
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
autoConnect: false,
});
await client.connect();
try {
await client.connect();
} catch (error) {
if (error instanceof AlreadyConnectedError) {
console.error("already connected");
}
}
await client.disconnect();
try {
await client.prompt({ sessionId: "s", prompt: [{ type: "text", text: "hi" }] });
} catch (error) {
if (error instanceof NotConnectedError) {
console.error("connect first");
}
}
```
## Session flow
```ts
const session = await client.newSession({
cwd: "/",
mcpServers: [],
metadata: {
agent: "mock",
title: "Demo Session",
variant: "high",
permissionMode: "ask",
},
});
const result = await client.prompt({
sessionId: session.sessionId,
prompt: [{ type: "text", text: "Summarize this repository." }],
});
console.log(result.stopReason);
```
Load, cancel, and runtime settings use ACP-aligned method names:
```ts
await client.loadSession({ sessionId: session.sessionId, cwd: "/", mcpServers: [] });
await client.cancel({ sessionId: session.sessionId });
await client.setSessionMode({ sessionId: session.sessionId, modeId: "default" });
await client.setSessionConfigOption({
sessionId: session.sessionId,
configId: "config-id-from-session",
value: "config-value-id",
});
```
## Extension helpers
Sandbox extensions are exposed as first-class methods:
```ts
const models = await client.listModels({ sessionId: session.sessionId });
console.log(models.currentModelId, models.availableModels.length);
await client.setMetadata(session.sessionId, {
title: "Renamed Session",
model: "mock",
permissionMode: "ask",
});
await client.detachSession(session.sessionId);
await client.terminateSession(session.sessionId);
```
## Event handling
Use `onEvent` to consume converted SDK events.
```ts
import { SandboxAgentClient, type AgentEvent } from "sandbox-agent";
const events: AgentEvent[] = [];
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
onEvent: (event) => {
events.push(event);
if (event.type === "sessionEnded") {
console.log("ended", event.notification.params.sessionId ?? event.notification.params.session_id);
}
if (event.type === "agentUnparsed") {
console.warn("unparsed", event.notification.params);
}
},
});
```
You can also handle raw session update notifications directly:
```ts
const client = new SandboxAgentClient({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
onSessionUpdate: (notification) => {
console.log(notification.update.sessionUpdate);
},
});
```
## Control + HTTP helpers
Agent/session and non-binary filesystem control helpers use ACP extension methods over `/v2/rpc`:
```ts
const health = await client.getHealth();
const agents = await client.listAgents();
await client.installAgent("codex", { reinstall: true });
const sessions = await client.listSessions();
const sessionInfo = await client.getSession(sessions.sessions[0].session_id);
```
These methods require an active ACP connection and throw `NotConnectedError` when disconnected.
Binary filesystem transfer intentionally remains HTTP:
- `readFsFile` -> `GET /v2/fs/file`
- `writeFsFile` -> `PUT /v2/fs/file`
- `uploadFsBatch` -> `POST /v2/fs/upload-batch`
Reason: these are Sandbox Agent host/runtime filesystem operations (not agent-specific ACP behavior), intentionally separate from ACP native `fs/read_text_file` / `fs/write_text_file`, and they may require streaming very large binary payloads that ACP JSON-RPC is not suited to transport efficiently.
ACP extension variants can exist in parallel for compatibility, but `SandboxAgentClient` should prefer the HTTP endpoints above by default.
## Error handling
All HTTP errors throw `SandboxAgentError`:
```ts
import { SandboxAgentError } from "sandbox-agent";
try {
await client.listAgents();
} catch (error) {
if (error instanceof SandboxAgentError) {
console.error(error.status, error.problem);
}
}
```
## Inspector URL
Build a URL to open the sandbox-agent Inspector UI with pre-filled connection settings:
```ts
import { buildInspectorUrl } from "sandbox-agent";
const url = buildInspectorUrl({
baseUrl: "https://your-sandbox-agent.example.com",
token: "optional-bearer-token",
headers: { "X-Custom-Header": "value" },
});
console.log(url);
// https://your-sandbox-agent.example.com/ui/?token=...&headers=...
```
Parameters:
- `baseUrl` (required): The sandbox-agent server URL
- `token` (optional): Bearer token for authentication
- `headers` (optional): Extra headers to pass to the server (JSON-encoded in the URL)
## Types
The SDK exports typed events and responses for the Sandbox layer:
```ts
import type {
AgentEvent,
AgentInfo,
HealthResponse,
SessionInfo,
SessionListResponse,
SessionTerminateResponse,
} from "sandbox-agent";
```
For low-level protocol transport details, see [ACP HTTP Client](/advanced/acp-http-client).

191
docs/security.mdx Normal file
View file

@ -0,0 +1,191 @@
---
title: "Security"
description: "Backend-first auth and access control patterns."
icon: "shield"
---
As covered in [Architecture](/architecture), run the Sandbox Agent client on your backend, not in the browser.
This keeps sandbox credentials private and gives you one place for authz, rate limiting, and audit logging.
## Auth model
Implement auth however it fits your stack (sessions, JWT, API keys, etc.), but enforce it before any sandbox-bound request.
Minimum checks:
- Authenticate the caller.
- Authorize access to the target workspace/sandbox/session.
- Apply request rate limits and request logging.
## Examples
### Rivet
<CodeGroup>
```ts Actor (server)
import { UserError, actor } from "rivetkit";
import { SandboxAgent } from "sandbox-agent";
type ConnParams = {
accessToken: string;
};
type WorkspaceClaims = {
sub: string;
workspaceId: string;
role: "owner" | "member" | "viewer";
};
async function verifyWorkspaceToken(
token: string,
workspaceId: string,
): Promise<WorkspaceClaims | null> {
// Validate JWT/session token here, then enforce workspace scope.
// Return null when invalid/expired/not a member.
if (!token) return null;
return { sub: "user_123", workspaceId, role: "member" };
}
export const workspace = actor({
state: {
events: [] as Array<{ userId: string; prompt: string; createdAt: number }>,
},
onBeforeConnect: async (c, params: ConnParams) => {
const claims = await verifyWorkspaceToken(params.accessToken, c.key[0]);
if (!claims) {
throw new UserError("Forbidden", { code: "forbidden" });
}
},
createConnState: async (c, params: ConnParams) => {
const claims = await verifyWorkspaceToken(params.accessToken, c.key[0]);
if (!claims) {
throw new UserError("Forbidden", { code: "forbidden" });
}
return {
userId: claims.sub,
role: claims.role,
workspaceId: claims.workspaceId,
};
},
actions: {
submitPrompt: async (c, prompt: string) => {
if (!c.conn) {
throw new UserError("Connection required", { code: "connection_required" });
}
if (c.conn.state.role === "viewer") {
throw new UserError("Insufficient permissions", { code: "forbidden" });
}
// Connect to Sandbox Agent from the actor (server-side only).
// Sandbox credentials never reach the client.
const sdk = await SandboxAgent.connect({
baseUrl: process.env.SANDBOX_URL!,
token: process.env.SANDBOX_TOKEN,
});
const session = await sdk.createSession({
agent: "claude",
sessionInit: { cwd: "/workspace" },
});
session.onEvent((event) => {
c.broadcast("session.event", {
userId: c.conn!.state.userId,
eventIndex: event.eventIndex,
sender: event.sender,
payload: event.payload,
});
});
const result = await session.prompt([
{ type: "text", text: prompt },
]);
c.state.events.push({
userId: c.conn.state.userId,
prompt,
createdAt: Date.now(),
});
return { stopReason: result.stopReason };
},
},
});
```
```ts Client (browser)
import { createClient } from "rivetkit/client";
import type { registry } from "./actors";
const client = createClient<typeof registry>({
endpoint: process.env.NEXT_PUBLIC_RIVET_ENDPOINT!,
});
const handle = client.workspace.getOrCreate(["ws_123"], {
params: { accessToken: userJwt },
});
const conn = handle.connect();
conn.on("session.event", (event) => {
console.log(event.sender, event.payload);
});
const result = await conn.submitPrompt("Plan a refactor for auth middleware.");
console.log(result.stopReason);
```
</CodeGroup>
Use [onBeforeConnect](https://rivet.dev/docs/actors/authentication), [connection params](https://rivet.dev/docs/actors/connections), and [actor keys](https://rivet.dev/docs/actors/keys) together so each actor enforces auth per workspace.
### Hono
```ts
import { Hono } from "hono";
import { bearerAuth } from "hono/bearer-auth";
const app = new Hono();
app.use("/sandbox/*", bearerAuth({ token: process.env.APP_API_TOKEN! }));
app.all("/sandbox/*", async (c) => {
const incoming = new URL(c.req.url);
const upstreamUrl = new URL(process.env.SANDBOX_URL!);
upstreamUrl.pathname = incoming.pathname.replace(/^\/sandbox/, "/v1");
upstreamUrl.search = incoming.search;
const headers = new Headers();
headers.set("authorization", `Bearer ${process.env.SANDBOX_TOKEN ?? ""}`);
const accept = c.req.header("accept");
if (accept) headers.set("accept", accept);
const contentType = c.req.header("content-type");
if (contentType) headers.set("content-type", contentType);
const body =
c.req.method === "POST" || c.req.method === "PUT" || c.req.method === "PATCH"
? await c.req.text()
: undefined;
const upstream = await fetch(upstreamUrl, {
method: c.req.method,
headers,
body,
});
return new Response(upstream.body, {
status: upstream.status,
headers: upstream.headers,
});
});
```

183
docs/session-persistence.mdx Normal file
View file

@ -0,0 +1,183 @@
---
title: "Persisting Sessions"
description: "Choose and configure session persistence for the TypeScript SDK."
icon: "database"
---
The TypeScript SDK uses a `SessionPersistDriver` to store session records and event history.
If you do not provide one, the SDK uses in-memory storage.
With persistence enabled, sessions can be restored after runtime/session loss. See [Session Restoration](/session-restoration).
Each driver stores:
- `SessionRecord` (`id`, `agent`, `agentSessionId`, `lastConnectionId`, `createdAt`, optional `destroyedAt`, optional `sessionInit`)
- `SessionEvent` (`id`, `eventIndex`, `sessionId`, `connectionId`, `sender`, `payload`, `createdAt`)
## Persistence drivers
### In-memory
Best for local dev and ephemeral workloads.
```ts
import { InMemorySessionPersistDriver, SandboxAgent } from "sandbox-agent";
const persist = new InMemorySessionPersistDriver({
maxSessions: 1024,
maxEventsPerSession: 500,
});
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
persist,
});
```
### Rivet
Recommended for sandbox orchestration with actor state.
```bash
npm install @sandbox-agent/persist-rivet@0.1.x
```
```ts
import { actor } from "rivetkit";
import { SandboxAgent } from "sandbox-agent";
import { RivetSessionPersistDriver, type RivetPersistState } from "@sandbox-agent/persist-rivet";
type PersistedState = RivetPersistState & {
sandboxId: string;
baseUrl: string;
};
export default actor({
createState: async () => {
return {
sandboxId: "sbx_123",
baseUrl: "http://127.0.0.1:2468",
} satisfies Partial<PersistedState>;
},
createVars: async (c) => {
const persist = new RivetSessionPersistDriver(c);
const sdk = await SandboxAgent.connect({
baseUrl: c.state.baseUrl,
persist,
});
const session = await sdk.resumeOrCreateSession({ id: "default", agent: "codex" });
const unsubscribe = session.onEvent((event) => {
c.broadcast("session.event", event);
});
return { sdk, session, unsubscribe };
},
actions: {
sendMessage: async (c, message: string) => {
await c.vars.session.prompt([{ type: "text", text: message }]);
},
},
onSleep: async (c) => {
c.vars.unsubscribe?.();
await c.vars.sdk.dispose();
},
});
```
### IndexedDB
Best for browser apps that should survive reloads.
```bash
npm install @sandbox-agent/persist-indexeddb@0.2.x
```
```ts
import { SandboxAgent } from "sandbox-agent";
import { IndexedDbSessionPersistDriver } from "@sandbox-agent/persist-indexeddb";
const persist = new IndexedDbSessionPersistDriver({
databaseName: "sandbox-agent-session-store",
});
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
persist,
});
```
### SQLite
Best for local/server Node apps that need durable storage without a DB server.
```bash
npm install @sandbox-agent/persist-sqlite@0.2.x
```
```ts
import { SandboxAgent } from "sandbox-agent";
import { SQLiteSessionPersistDriver } from "@sandbox-agent/persist-sqlite";
const persist = new SQLiteSessionPersistDriver({
filename: "./sandbox-agent.db",
});
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
persist,
});
```
### Postgres
Use when you already run Postgres and want shared relational storage.
```bash
npm install @sandbox-agent/persist-postgres@0.2.x
```
```ts
import { SandboxAgent } from "sandbox-agent";
import { PostgresSessionPersistDriver } from "@sandbox-agent/persist-postgres";
const persist = new PostgresSessionPersistDriver({
connectionString: process.env.DATABASE_URL,
schema: "public",
});
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
persist,
});
```
### Custom driver
Implement `SessionPersistDriver` for custom backends.
```ts
import type { SessionPersistDriver } from "sandbox-agent";
class MyDriver implements SessionPersistDriver {
async getSession(id) { return null; }
async listSessions(request) { return { items: [] }; }
async updateSession(session) {}
async listEvents(request) { return { items: [] }; }
async insertEvent(event) {}
}
```
## Replay controls
`SandboxAgent.connect(...)` supports:
- `replayMaxEvents` (default `50`)
- `replayMaxChars` (default `12000`)
These cap replay size when restoring sessions.
## Related docs
- [SDK Overview](/sdk-overview)
- [Session Restoration](/session-restoration)

33
docs/session-restoration.mdx Normal file
View file

@ -0,0 +1,33 @@
---
title: "Session Restoration"
description: "How the TypeScript SDK restores sessions after connection/runtime loss."
---
Sandbox Agent automatically restores stale sessions when live session state is no longer available.
This is driven by the configured `SessionPersistDriver` (`inMemory`, IndexedDB, SQLite, Postgres, or custom).
## How Auto-Restore Works
When you call `session.prompt(...)` (or `resumeSession(...)`) and the saved session points to a stale connection, the SDK:
1. Recreates a fresh session for the same local session id.
2. Rebinds the local session to the new runtime session id.
3. Replays recent persisted events into the next prompt as context.
This happens automatically; you do not need to manually rebuild the session.
## Replay Limits
Replay payload size is capped by:
- `replayMaxEvents` (default `50`)
- `replayMaxChars` (default `12000`)
These controls limit prompt growth during restore while preserving recent context.
## Related Docs
- [SDK Overview](/sdk-overview)
- [Persisting Sessions](/session-persistence)
- [Agent Sessions](/agent-sessions)

388
docs/session-transcript-schema.mdx
View file

@ -1,388 +0,0 @@
---
title: "Session Transcript Schema"
description: "Universal event schema for session transcripts across all agents."
---
Each coding agent outputs events in its own native format. The sandbox-agent converts these into a universal event schema, giving you a consistent session transcript regardless of which agent you use.
The schema is defined in [OpenAPI format](https://github.com/rivet-dev/sandbox-agent/blob/main/docs/openapi.json). See the [HTTP API Reference](/api-reference) for endpoint documentation.
## Coverage Matrix
This table shows which agent feature coverage appears in the universal event stream. All agents retain their full native feature coverage—this only reflects what's normalized into the schema.
| Feature | Claude | Codex | OpenCode | Amp | Pi (RPC) |
|--------------------|:------:|:-----:|:------------:|:------------:|:------------:|
| Stability | Stable | Stable| Experimental | Experimental | Experimental |
| Text Messages | ✓ | ✓ | ✓ | ✓ | ✓ |
| Tool Calls | ✓ | ✓ | ✓ | ✓ | ✓ |
| Tool Results | ✓ | ✓ | ✓ | ✓ | ✓ |
| Questions (HITL) | ✓ | | ✓ | | |
| Permissions (HITL) | ✓ | ✓ | ✓ | - | |
| Images | - | ✓ | ✓ | - | ✓ |
| File Attachments | - | ✓ | ✓ | - | |
| Session Lifecycle | - | ✓ | ✓ | - | |
| Error Events | - | ✓ | ✓ | ✓ | ✓ |
| Reasoning/Thinking | - | ✓ | - | - | ✓ |
| Command Execution | - | ✓ | - | - | |
| File Changes | - | ✓ | - | - | |
| MCP Tools | ✓ | ✓ | ✓ | ✓ | |
| Streaming Deltas | ✓ | ✓ | ✓ | - | ✓ |
| Variants | | ✓ | ✓ | ✓ | ✓ |
Agents: [Claude Code](https://docs.anthropic.com/en/docs/agents-and-tools/claude-code/overview) · [Codex](https://github.com/openai/codex) · [OpenCode](https://github.com/opencode-ai/opencode) · [Amp](https://ampcode.com) · [Pi](https://buildwithpi.ai/pi-cli)
- ✓ = Appears in session events
- \- = Agent supports natively, schema conversion coming soon
- (blank) = Not supported by agent
- Pi runtime model is router-managed per-session RPC (`pi --mode rpc`); it does not use generic subprocess streaming.
<AccordionGroup>
<Accordion title="Text Messages">
Basic message exchange between user and assistant.
</Accordion>
<Accordion title="Tool Calls & Results">
Visibility into tool invocations (file reads, command execution, etc.) and their results. When not natively supported, tool activity is embedded in message content.
</Accordion>
<Accordion title="Questions (HITL)">
Interactive questions the agent asks the user. Emits `question.requested` and `question.resolved` events.
</Accordion>
<Accordion title="Permissions (HITL)">
Permission requests for sensitive operations. Emits `permission.requested` and `permission.resolved` events.
</Accordion>
<Accordion title="Images">
Support for image attachments in messages.
</Accordion>
<Accordion title="File Attachments">
Support for file attachments in messages.
</Accordion>
<Accordion title="Session Lifecycle">
Native `session.started` and `session.ended` events. When not supported, the daemon emits synthetic lifecycle events.
</Accordion>
<Accordion title="Error Events">
Structured error events for runtime failures.
</Accordion>
<Accordion title="Reasoning/Thinking">
Extended thinking or reasoning content with visibility controls.
</Accordion>
<Accordion title="Command Execution">
Detailed command execution events with stdout/stderr.
</Accordion>
<Accordion title="File Changes">
Structured file modification events with diffs.
</Accordion>
<Accordion title="MCP Tools">
Model Context Protocol tool support.
</Accordion>
<Accordion title="Streaming Deltas">
Native streaming of content deltas. When not supported, the daemon emits a single synthetic delta before `item.completed`.
</Accordion>
<Accordion title="Variants">
Model variants such as reasoning effort or depth. Agents may expose different variant sets per model.
</Accordion>
</AccordionGroup>
Want support for another agent? [Open an issue](https://github.com/rivet-dev/sandbox-agent/issues/new) to request it.
## UniversalEvent
Every event from the API is wrapped in a `UniversalEvent` envelope.
| Field | Type | Description |
|-------|------|-------------|
| `event_id` | string | Unique identifier for this event |
| `sequence` | integer | Monotonic sequence number within the session (starts at 1) |
| `time` | string | RFC3339 timestamp |
| `session_id` | string | Daemon-generated session identifier |
| `native_session_id` | string? | Provider-native session/thread identifier (e.g., Codex `threadId`, OpenCode `sessionID`) |
| `source` | string | Event origin: `agent` (native) or `daemon` (synthetic) |
| `synthetic` | boolean | Whether this event was generated by the daemon to fill gaps |
| `type` | string | Event type (see [Event Types](#event-types)) |
| `data` | object | Event-specific payload |
| `raw` | any? | Original provider payload (only when `include_raw=true`) |
```json
{
"event_id": "evt_abc123",
"sequence": 1,
"time": "2025-01-28T12:00:00Z",
"session_id": "my-session",
"native_session_id": "thread_xyz",
"source": "agent",
"synthetic": false,
"type": "item.completed",
"data": { ... }
}
```
## Event Types
### Session Lifecycle
| Type | Description | Data |
|------|-------------|------|
| `session.started` | Session has started | `{ metadata?: any }` |
| `session.ended` | Session has ended | `{ reason, terminated_by, message?, exit_code? }` |
### Turn Lifecycle
| Type | Description | Data |
|------|-------------|------|
| `turn.started` | Turn has started | `{ phase: "started", turn_id?, metadata? }` |
| `turn.ended` | Turn has ended | `{ phase: "ended", turn_id?, metadata? }` |
**SessionEndedData**
| Field | Type | Values |
|-------|------|--------|
| `reason` | string | `completed`, `error`, `terminated` |
| `terminated_by` | string | `agent`, `daemon` |
| `message` | string? | Error message (only present when reason is `error`) |
| `exit_code` | int? | Process exit code (only present when reason is `error`) |
| `stderr` | StderrOutput? | Structured stderr output (only present when reason is `error`) |
**StderrOutput**
| Field | Type | Description |
|-------|------|-------------|
| `head` | string? | First 20 lines of stderr (if truncated) or full stderr (if not truncated) |
| `tail` | string? | Last 50 lines of stderr (only present if truncated) |
| `truncated` | boolean | Whether the output was truncated |
| `total_lines` | int? | Total number of lines in stderr |
### Item Lifecycle
| Type | Description | Data |
|------|-------------|------|
| `item.started` | Item creation | `{ item }` |
| `item.delta` | Streaming content delta | `{ item_id, native_item_id?, delta }` |
| `item.completed` | Item finalized | `{ item }` |
Items follow a consistent lifecycle: `item.started` → `item.delta` (0 or more) → `item.completed`.
### HITL (Human-in-the-Loop)
| Type | Description | Data |
|------|-------------|------|
| `permission.requested` | Permission request pending | `{ permission_id, action, status, metadata? }` |
| `permission.resolved` | Permission decision recorded | `{ permission_id, action, status, metadata? }` |
| `question.requested` | Question pending user input | `{ question_id, prompt, options, status }` |
| `question.resolved` | Question answered or rejected | `{ question_id, prompt, options, status, response? }` |
**PermissionEventData**
| Field | Type | Description |
|-------|------|-------------|
| `permission_id` | string | Identifier for the permission request |
| `action` | string | What the agent wants to do |
| `status` | string | `requested`, `accept`, `accept_for_session`, `reject` |
| `metadata` | any? | Additional context |
**QuestionEventData**
| Field | Type | Description |
|-------|------|-------------|
| `question_id` | string | Identifier for the question |
| `prompt` | string | Question text |
| `options` | string[] | Available answer options |
| `status` | string | `requested`, `answered`, `rejected` |
| `response` | string? | Selected answer (when resolved) |
### Errors
| Type | Description | Data |
|------|-------------|------|
| `error` | Runtime error | `{ message, code?, details? }` |
| `agent.unparsed` | Parse failure | `{ error, location, raw_hash? }` |
The `agent.unparsed` event indicates the daemon failed to parse an agent payload. This should be treated as a bug.
## UniversalItem
Items represent discrete units of content within a session.
| Field | Type | Description |
|-------|------|-------------|
| `item_id` | string | Daemon-generated identifier |
| `native_item_id` | string? | Provider-native item/message identifier |
| `parent_id` | string? | Parent item ID (e.g., tool call/result parented to a message) |
| `kind` | string | Item category (see below) |
| `role` | string? | Actor role for message items |
| `status` | string | Lifecycle status |
| `content` | ContentPart[] | Ordered list of content parts |
### ItemKind
| Value | Description |
|-------|-------------|
| `message` | User or assistant message |
| `tool_call` | Tool invocation |
| `tool_result` | Tool execution result |
| `system` | System message |
| `status` | Status update |
| `unknown` | Unrecognized item type |
### ItemRole
| Value | Description |
|-------|-------------|
| `user` | User message |
| `assistant` | Assistant response |
| `system` | System prompt |
| `tool` | Tool-related message |
### ItemStatus
| Value | Description |
|-------|-------------|
| `in_progress` | Item is streaming or pending |
| `completed` | Item is finalized |
| `failed` | Item execution failed |
## Content Parts
The `content` array contains typed parts that make up an item's payload.
### text
Plain text content.
```json
{ "type": "text", "text": "Hello, world!" }
```
### json
Structured JSON content.
```json
{ "type": "json", "json": { "key": "value" } }
```
### tool_call
Tool invocation.
| Field | Type | Description |
|-------|------|-------------|
| `name` | string | Tool name |
| `arguments` | string | JSON-encoded arguments |
| `call_id` | string | Unique call identifier |
```json
{
"type": "tool_call",
"name": "read_file",
"arguments": "{\"path\": \"/src/main.ts\"}",
"call_id": "call_abc123"
}
```
### tool_result
Tool execution result.
| Field | Type | Description |
|-------|------|-------------|
| `call_id` | string | Matching call identifier |
| `output` | string | Tool output |
```json
{
"type": "tool_result",
"call_id": "call_abc123",
"output": "File contents here..."
}
```
### file_ref
File reference with optional diff.
| Field | Type | Description |
|-------|------|-------------|
| `path` | string | File path |
| `action` | string | `read`, `write`, `patch` |
| `diff` | string? | Unified diff (for patches) |
```json
{
"type": "file_ref",
"path": "/src/main.ts",
"action": "write",
"diff": "@@ -1,3 +1,4 @@\n+import { foo } from 'bar';"
}
```
### image
Image reference.
| Field | Type | Description |
|-------|------|-------------|
| `path` | string | Image file path |
| `mime` | string? | MIME type |
```json
{ "type": "image", "path": "/tmp/screenshot.png", "mime": "image/png" }
```
### reasoning
Model reasoning/thinking content.
| Field | Type | Description |
|-------|------|-------------|
| `text` | string | Reasoning text |
| `visibility` | string | `public` or `private` |
```json
{ "type": "reasoning", "text": "Let me think about this...", "visibility": "public" }
```
### status
Status indicator.
| Field | Type | Description |
|-------|------|-------------|
| `label` | string | Status label |
| `detail` | string? | Additional detail |
```json
{ "type": "status", "label": "Running tests", "detail": "3 of 10 passed" }
```
## Source & Synthetics
### EventSource
The `source` field indicates who emitted the event:
| Value | Description |
|-------|-------------|
| `agent` | Native event from the agent |
| `daemon` | Synthetic event generated by the daemon |
### Synthetic Events
The daemon emits synthetic events (`synthetic: true`, `source: "daemon"`) to provide a consistent event stream across all agents. Common synthetics:
| Synthetic | When |
|-----------|------|
| `session.started` | Agent doesn't emit explicit session start |
| `session.ended` | Agent doesn't emit explicit session end |
| `turn.started` | Agent doesn't emit explicit turn start |
| `turn.ended` | Agent doesn't emit explicit turn end |
| `item.started` | Agent doesn't emit item start events |
| `item.delta` | Agent doesn't stream deltas natively |
| `question.*` | Claude Code plan mode (from ExitPlanMode tool) |
### Raw Payloads
Pass `include_raw=true` to event endpoints to receive the original agent payload in the `raw` field. Useful for debugging or accessing agent-specific data not in the universal schema.
```typescript
const events = await client.getEvents("my-session", { includeRaw: true });
// events[0].raw contains the original agent payload
```

107
docs/skills-config.mdx
View file

@ -1,88 +1,81 @@
---
title: "Skills"
description: "Auto-load skills into agent sessions."
description: "Configure skill sources for agent sessions."
sidebarTitle: "Skills"
icon: "sparkles"
---
Skills are local instruction bundles stored in `SKILL.md` files. Sandbox Agent can fetch, discover, and link skill directories into agent-specific skill paths at session start using the `skills.sources` field. The format is fully compatible with [skills.sh](https://skills.sh).
Skills are local instruction bundles stored in `SKILL.md` files.
## Session Config
## Configuring skills
Pass `skills.sources` when creating a session to load skills from GitHub repos, local paths, or git URLs.
Use `setSkillsConfig` / `getSkillsConfig` / `deleteSkillsConfig` to manage skill source config by directory + skill name.
<CodeGroup>
```ts
import { SandboxAgent } from "sandbox-agent";
```ts TypeScript
import { SandboxAgentClient } from "sandbox-agent";
const client = new SandboxAgentClient({
const sdk = await SandboxAgent.connect({
baseUrl: "http://127.0.0.1:2468",
token: process.env.SANDBOX_TOKEN,
agent: "mock",
});
await client.createSession("claude-skills", {
agent: "claude",
skills: {
// Add a skill
await sdk.setSkillsConfig(
{
directory: "/workspace",
skillName: "default",
},
{
sources: [
{ type: "github", source: "rivet-dev/skills", skills: ["sandbox-agent"] },
{ type: "local", source: "/workspace/my-custom-skill" },
],
},
);
// Create a session using the configured skills
const session = await sdk.createSession({
agent: "claude",
sessionInit: {
cwd: "/workspace",
},
});
await session.prompt([
{ type: "text", text: "Use available skills to help with this task." },
]);
// List skills
const config = await sdk.getSkillsConfig({
directory: "/workspace",
skillName: "default",
});
console.log(config.sources.length);
// Delete skill
await sdk.deleteSkillsConfig({
directory: "/workspace",
skillName: "default",
});
```
```bash cURL
curl -X POST "http://127.0.0.1:2468/v1/sessions/claude-skills" \
-H "Authorization: Bearer $SANDBOX_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"agent": "claude",
"skills": {
"sources": [
{ "type": "github", "source": "rivet-dev/skills", "skills": ["sandbox-agent"] },
{ "type": "local", "source": "/workspace/my-custom-skill" }
]
}
}'
```
## Skill sources
</CodeGroup>
Each skill directory must contain `SKILL.md`. See [Skill authoring best practices](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices) for tips on writing effective skills.
## Skill Sources
Each entry in `skills.sources` describes where to find skills. Three source types are supported:
Each `skills.sources` entry describes where to find skills.
| Type | `source` value | Example |
|------|---------------|---------|
| `github` | `owner/repo` | `"rivet-dev/skills"` |
| `local` | Filesystem path | `"/workspace/my-skill"` |
| `git` | Git clone URL | `"https://git.example.com/skills.git"` |
| `local` | filesystem path | `"/workspace/my-skill"` |
| `git` | git clone URL | `"https://git.example.com/skills.git"` |
### Optional fields
Optional fields:
- **`skills`** — Array of skill directory names to include. When omitted, all discovered skills are installed.
- **`ref`** — Branch, tag, or commit to check out (default: HEAD). Applies to `github` and `git` types.
- **`subpath`** — Subdirectory within the repo to search for skills.
- `skills`: subset of skill directory names to include
- `ref`: branch/tag/commit (for `github` and `git`)
- `subpath`: subdirectory within repo to scan
## Custom Skills
## Custom skills
To write, upload, and configure your own skills inside the sandbox, see [Custom Tools](/custom-tools).
## Advanced
### Discovery logic
After resolving a source to a local directory (cloning if needed), Sandbox Agent discovers skills by:
1. Checking if the directory itself contains `SKILL.md`.
2. Scanning `skills/` subdirectory for child directories containing `SKILL.md`.
3. Scanning immediate children of the directory for `SKILL.md`.
Discovered skills are symlinked into project-local skill roots (`.claude/skills/<name>`, `.agents/skills/<name>`, `.opencode/skill/<name>`).
### Caching
GitHub sources are downloaded as zip archives and git sources are cloned to `~/.sandbox-agent/skills-cache/` and updated on subsequent session creations. GitHub sources do not require `git` to be installed.

14
examples/CLAUDE.md
View file

@ -6,16 +6,16 @@
- Do not bind mount host files or host directories into Docker example containers.
- If an example needs tools, skills, or MCP servers, install them inside the container during setup.
## Testing Examples (ACP v2)
## Testing Examples (ACP v1)
Examples should be validated against v2 endpoints:
Examples should be validated against v1 endpoints:
1. Start the example: `SANDBOX_AGENT_DEV=1 pnpm start`
2. Create an ACP client by POSTing `initialize` to `/v2/rpc` with `x-acp-agent: mock` (or another installed agent).
3. Capture `x-acp-connection-id` from the response headers.
4. Open SSE stream: `GET /v2/rpc` with `x-acp-connection-id`.
5. Send `session/new` then `session/prompt` via `POST /v2/rpc` with the same connection id.
6. Close connection via `DELETE /v2/rpc` with `x-acp-connection-id`.
2. Pick a server id, for example `example-smoke`.
3. Create ACP transport by POSTing `initialize` to `/v1/acp/example-smoke?agent=mock` (or another installed agent).
4. Open SSE stream: `GET /v1/acp/example-smoke`.
5. Send `session/new` then `session/prompt` via `POST /v1/acp/example-smoke`.
6. Close connection via `DELETE /v1/acp/example-smoke`.
v1 reminder:

2
examples/cloudflare/Dockerfile
View file

@ -1,7 +1,7 @@
FROM cloudflare/sandbox:0.7.0
# Install sandbox-agent
RUN curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh
RUN curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh
# Pre-install agents
RUN sandbox-agent install-agent claude && \

8
examples/daytona/src/daytona-with-snapshot.ts
View file

@ -1,6 +1,6 @@
import { Daytona, Image } from "@daytonaio/sdk";
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId, waitForHealth } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl, waitForHealth } from "@sandbox-agent/example-shared";
const daytona = new Daytona();
@ -13,7 +13,7 @@ if (process.env.OPENAI_API_KEY)
// Build a custom image with sandbox-agent pre-installed (slower first run, faster subsequent runs)
const image = Image.base("ubuntu:22.04").runCommands(
"apt-get update && apt-get install -y curl ca-certificates",
"curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh",
"curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh",
);
console.log("Creating Daytona sandbox (first run builds the base image and may take a few minutes, subsequent runs are fast)...");
@ -29,8 +29,8 @@ console.log("Waiting for server...");
await waitForHealth({ baseUrl });
const client = await SandboxAgent.connect({ baseUrl });
const sessionId = generateSessionId();
await client.createSession(sessionId, { agent: detectAgent() });
const session = await client.createSession({ agent: detectAgent(), sessionInit: { cwd: "/root" } });
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(" Press Ctrl+C to stop.");

8
examples/daytona/src/index.ts
View file

@ -1,6 +1,6 @@
import { Daytona } from "@daytonaio/sdk";
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId, waitForHealth } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl, waitForHealth } from "@sandbox-agent/example-shared";
const daytona = new Daytona();
@ -17,7 +17,7 @@ const sandbox = await daytona.create({ envVars, autoStopInterval: 0 });
// Install sandbox-agent and start server
console.log("Installing sandbox-agent...");
await sandbox.process.executeCommand(
"curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh",
"curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh",
);
await sandbox.process.executeCommand(
@ -30,8 +30,8 @@ console.log("Waiting for server...");
await waitForHealth({ baseUrl });
const client = await SandboxAgent.connect({ baseUrl });
const sessionId = generateSessionId();
await client.createSession(sessionId, { agent: detectAgent() });
const session = await client.createSession({ agent: detectAgent(), sessionInit: { cwd: "/root" } });
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(" Press Ctrl+C to stop.");

8
examples/docker/src/index.ts
View file

@ -1,6 +1,6 @@
import Docker from "dockerode";
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId, waitForHealth } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl, waitForHealth } from "@sandbox-agent/example-shared";
const IMAGE = "alpine:latest";
const PORT = 3000;
@ -25,7 +25,7 @@ const container = await docker.createContainer({
Image: IMAGE,
Cmd: ["sh", "-c", [
"apk add --no-cache curl ca-certificates libstdc++ libgcc bash",
"curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh",
"curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh",
"sandbox-agent install-agent claude",
"sandbox-agent install-agent codex",
`sandbox-agent server --no-token --host 0.0.0.0 --port ${PORT}`,
@ -46,8 +46,8 @@ const baseUrl = `http://127.0.0.1:${PORT}`;
await waitForHealth({ baseUrl });
const client = await SandboxAgent.connect({ baseUrl });
const sessionId = generateSessionId();
await client.createSession(sessionId, { agent: detectAgent() });
const session = await client.createSession({ agent: detectAgent(), sessionInit: { cwd: "/root" } });
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(" Press Ctrl+C to stop.");

8
examples/e2b/src/index.ts
View file

@ -1,6 +1,6 @@
import { Sandbox } from "@e2b/code-interpreter";
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId, waitForHealth } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl, waitForHealth } from "@sandbox-agent/example-shared";
const envs: Record<string, string> = {};
if (process.env.ANTHROPIC_API_KEY) envs.ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;
@ -16,7 +16,7 @@ const run = async (cmd: string) => {
};
console.log("Installing sandbox-agent...");
await run("curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh");
await run("curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh");
console.log("Installing agents...");
await run("sandbox-agent install-agent claude");
@ -31,8 +31,8 @@ console.log("Waiting for server...");
await waitForHealth({ baseUrl });
const client = await SandboxAgent.connect({ baseUrl });
const sessionId = generateSessionId();
await client.createSession(sessionId, { agent: detectAgent() });
const session = await client.createSession({ agent: detectAgent(), sessionInit: { cwd: "/root" } });
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(" Press Ctrl+C to stop.");

6
examples/file-system/src/index.ts
View file

@ -1,5 +1,5 @@
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl } from "@sandbox-agent/example-shared";
import { startDockerSandbox } from "@sandbox-agent/example-shared/docker";
import * as tar from "tar";
import fs from "node:fs";
@ -47,8 +47,8 @@ const readmeText = new TextDecoder().decode(readmeBytes);
console.log(` README.md content: ${readmeText.trim()}`);
console.log("Creating session...");
const sessionId = generateSessionId();
await client.createSession(sessionId, { agent: detectAgent() });
const session = await client.createSession({ agent: detectAgent(), sessionInit: { cwd: "/opt/my-project" } });
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(' Try: "read the README in /opt/my-project"');
console.log(" Press Ctrl+C to stop.");

19
examples/mcp-custom-tool/src/index.ts
View file

@ -1,5 +1,5 @@
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl } from "@sandbox-agent/example-shared";
import { startDockerSandbox } from "@sandbox-agent/example-shared/docker";
import fs from "node:fs";
import path from "node:path";
@ -31,16 +31,19 @@ console.log(` Written: ${written.path} (${written.bytesWritten} bytes)`);
// Create a session with the uploaded MCP server as a local command.
console.log("Creating session with custom MCP tool...");
const sessionId = generateSessionId();
await client.createSession(sessionId, {
const session = await client.createSession({
agent: detectAgent(),
mcp: {
customTools: {
type: "local",
command: ["node", "/opt/mcp/custom-tools/mcp-server.cjs"],
},
sessionInit: {
cwd: "/root",
mcpServers: [{
name: "customTools",
command: "node",
args: ["/opt/mcp/custom-tools/mcp-server.cjs"],
env: [],
}],
},
});
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(' Try: "generate a random number between 1 and 100"');
console.log(" Press Ctrl+C to stop.");

20
examples/mcp/src/index.ts
View file

@ -1,5 +1,5 @@
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl } from "@sandbox-agent/example-shared";
import { startDockerSandbox } from "@sandbox-agent/example-shared/docker";
console.log("Starting sandbox...");
@ -12,17 +12,19 @@ const { baseUrl, cleanup } = await startDockerSandbox({
console.log("Creating session with everything MCP server...");
const client = await SandboxAgent.connect({ baseUrl });
const sessionId = generateSessionId();
await client.createSession(sessionId, {
const session = await client.createSession({
agent: detectAgent(),
mcp: {
everything: {
type: "local",
command: ["mcp-server-everything"],
timeoutMs: 10000,
},
sessionInit: {
cwd: "/root",
mcpServers: [{
name: "everything",
command: "mcp-server-everything",
args: [],
env: [],
}],
},
});
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(' Try: "generate a random number between 1 and 100"');
console.log(" Press Ctrl+C to stop.");

9
examples/mock-acp-agent/README.md Normal file
View file

@ -0,0 +1,9 @@
# @sandbox-agent/mock-acp-agent
Minimal newline-delimited ACP JSON-RPC mock agent.
Behavior:
- Echoes every inbound message as `mock/echo` notification.
- For requests (`method` + `id`), returns `result.echoed` payload.
- For `mock/ask_client`, emits an agent-initiated `mock/request` before response.
- For responses from client (`id` without `method`), emits `mock/client_response` notification.

24
examples/mock-acp-agent/package.json Normal file
View file

@ -0,0 +1,24 @@
{
"name": "@sandbox-agent/mock-acp-agent",
"version": "0.1.0",
"private": false,
"type": "module",
"description": "Mock ACP agent for adapter integration testing",
"license": "Apache-2.0",
"main": "./dist/index.js",
"exports": {
".": "./dist/index.js"
},
"files": [
"dist"
],
"scripts": {
"build": "tsc -p tsconfig.build.json",
"typecheck": "tsc --noEmit",
"start": "node ./dist/index.js"
},
"devDependencies": {
"@types/node": "latest",
"typescript": "latest"
}
}

100
examples/mock-acp-agent/src/index.ts Normal file
View file

@ -0,0 +1,100 @@
import { createInterface } from "node:readline";
interface JsonRpcRequest {
jsonrpc?: unknown;
id?: unknown;
method?: unknown;
params?: unknown;
result?: unknown;
error?: unknown;
}
let outboundRequestSeq = 0;
function writeMessage(payload: unknown): void {
process.stdout.write(`${JSON.stringify(payload)}\n`);
}
function echoNotification(message: unknown): void {
writeMessage({
jsonrpc: "2.0",
method: "mock/echo",
params: {
message,
},
});
}
function handleMessage(raw: string): void {
if (!raw.trim()) {
return;
}
let msg: JsonRpcRequest;
try {
msg = JSON.parse(raw) as JsonRpcRequest;
} catch (error) {
writeMessage({
jsonrpc: "2.0",
method: "mock/parse_error",
params: {
error: error instanceof Error ? error.message : String(error),
raw,
},
});
return;
}
echoNotification(msg);
const hasMethod = typeof msg.method === "string";
const hasId = msg.id !== undefined;
if (hasMethod && hasId) {
if (msg.method === "mock/ask_client") {
outboundRequestSeq += 1;
writeMessage({
jsonrpc: "2.0",
id: `agent-req-${outboundRequestSeq}`,
method: "mock/request",
params: {
prompt: "please respond",
},
});
}
writeMessage({
jsonrpc: "2.0",
id: msg.id,
result: {
echoed: msg,
},
});
return;
}
if (!hasMethod && hasId) {
writeMessage({
jsonrpc: "2.0",
method: "mock/client_response",
params: {
id: msg.id,
result: msg.result ?? null,
error: msg.error ?? null,
},
});
}
}
const rl = createInterface({
input: process.stdin,
crlfDelay: Infinity,
});
rl.on("line", (line) => {
handleMessage(line);
});
rl.on("close", () => {
process.exit(0);
});

11
examples/mock-acp-agent/tsconfig.build.json Normal file
View file

@ -0,0 +1,11 @@
{
"extends": "./tsconfig.json",
"compilerOptions": {
"allowImportingTsExtensions": false,
"noEmit": false,
"declaration": true,
"outDir": "./dist",
"rootDir": "./src"
},
"include": ["src/**/*"]
}

17
examples/mock-acp-agent/tsconfig.json Normal file
View file

@ -0,0 +1,17 @@
{
"compilerOptions": {
"target": "ES2022",
"lib": ["ES2022"],
"module": "ES2022",
"moduleResolution": "Bundler",
"allowImportingTsExtensions": true,
"noEmit": true,
"esModuleInterop": true,
"strict": true,
"skipLibCheck": true,
"resolveJsonModule": true,
"types": ["node"]
},
"include": ["src/**/*"],
"exclude": ["node_modules"]
}

6
examples/shared/Dockerfile.dev
View file

@ -6,9 +6,11 @@ WORKDIR /build
COPY package.json pnpm-lock.yaml pnpm-workspace.yaml ./
# Copy packages needed for the inspector build chain:
# inspector -> sandbox-agent SDK -> cli-shared
# inspector -> sandbox-agent SDK -> acp-http-client, cli-shared, persist-indexeddb
COPY sdks/typescript/ sdks/typescript/
COPY sdks/acp-http-client/ sdks/acp-http-client/
COPY sdks/cli-shared/ sdks/cli-shared/
COPY sdks/persist-indexeddb/ sdks/persist-indexeddb/
COPY frontend/packages/inspector/ frontend/packages/inspector/
COPY docs/openapi.json docs/
@ -16,6 +18,7 @@ COPY docs/openapi.json docs/
# but not needed for the inspector build (avoids install errors).
RUN set -e; for dir in \
sdks/cli sdks/gigacode \
sdks/persist-postgres sdks/persist-sqlite sdks/persist-rivet \
resources/agent-schemas resources/vercel-ai-sdk-schemas \
scripts/release scripts/sandbox-testing \
examples/shared examples/docker examples/e2b examples/vercel \
@ -44,6 +47,7 @@ COPY Cargo.toml Cargo.lock ./
COPY server/ ./server/
COPY gigacode/ ./gigacode/
COPY resources/agent-schemas/artifacts/ ./resources/agent-schemas/artifacts/
COPY scripts/agent-configs/ ./scripts/agent-configs/
COPY --from=frontend /build/frontend/packages/inspector/dist/ ./frontend/packages/inspector/dist/
RUN --mount=type=cache,target=/usr/local/cargo/registry \
--mount=type=cache,target=/usr/local/cargo/git \

20
examples/skills-custom-tool/src/index.ts
View file

@ -1,5 +1,5 @@
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl } from "@sandbox-agent/example-shared";
import { startDockerSandbox } from "@sandbox-agent/example-shared/docker";
import fs from "node:fs";
import path from "node:path";
@ -36,15 +36,17 @@ const skillResult = await client.writeFsFile(
);
console.log(` Skill: ${skillResult.path} (${skillResult.bytesWritten} bytes)`);
// Create a session with the uploaded skill as a local source.
// Configure the uploaded skill.
console.log("Configuring custom skill...");
await client.setSkillsConfig(
{ directory: "/", skillName: "random-number" },
{ sources: [{ type: "local", source: "/opt/skills/random-number" }] },
);
// Create a session.
console.log("Creating session with custom skill...");
const sessionId = generateSessionId();
await client.createSession(sessionId, {
agent: detectAgent(),
skills: {
sources: [{ type: "local", source: "/opt/skills/random-number" }],
},
});
const session = await client.createSession({ agent: detectAgent(), sessionInit: { cwd: "/root" } });
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(' Try: "generate a random number between 1 and 100"');
console.log(" Press Ctrl+C to stop.");

21
examples/skills/src/index.ts
View file

@ -1,5 +1,5 @@
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl } from "@sandbox-agent/example-shared";
import { startDockerSandbox } from "@sandbox-agent/example-shared/docker";
console.log("Starting sandbox...");
@ -7,17 +7,16 @@ const { baseUrl, cleanup } = await startDockerSandbox({
port: 3001,
});
console.log("Creating session with skill source...");
console.log("Configuring skill source...");
const client = await SandboxAgent.connect({ baseUrl });
const sessionId = generateSessionId();
await client.createSession(sessionId, {
agent: detectAgent(),
skills: {
sources: [
{ type: "github", source: "rivet-dev/skills", skills: ["sandbox-agent"] },
],
},
});
await client.setSkillsConfig(
{ directory: "/", skillName: "rivet-dev-skills" },
{ sources: [{ type: "github", source: "rivet-dev/skills", skills: ["sandbox-agent"] }] },
);
console.log("Creating session...");
const session = await client.createSession({ agent: detectAgent(), sessionInit: { cwd: "/root" } });
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(' Try: "How do I start sandbox-agent?"');
console.log(" Press Ctrl+C to stop.");

8
examples/vercel/src/index.ts
View file

@ -1,6 +1,6 @@
import { Sandbox } from "@vercel/sandbox";
import { SandboxAgent } from "sandbox-agent";
import { detectAgent, buildInspectorUrl, generateSessionId, waitForHealth } from "@sandbox-agent/example-shared";
import { detectAgent, buildInspectorUrl, waitForHealth } from "@sandbox-agent/example-shared";
const envs: Record<string, string> = {};
if (process.env.ANTHROPIC_API_KEY) envs.ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;
@ -22,7 +22,7 @@ const run = async (cmd: string, args: string[] = []) => {
};
console.log("Installing sandbox-agent...");
await run("sh", ["-c", "curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh"]);
await run("sh", ["-c", "curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh"]);
console.log("Installing agents...");
await run("sandbox-agent", ["install-agent", "claude"]);
@ -42,8 +42,8 @@ console.log("Waiting for server...");
await waitForHealth({ baseUrl });
const client = await SandboxAgent.connect({ baseUrl });
const sessionId = generateSessionId();
await client.createSession(sessionId, { agent: detectAgent() });
const session = await client.createSession({ agent: detectAgent(), sessionInit: { cwd: "/root" } });
const sessionId = session.id;
console.log(` UI: ${buildInspectorUrl({ baseUrl, sessionId })}`);
console.log(" Press Ctrl+C to stop.");

2
frontend/CLAUDE.md
View file

@ -3,7 +3,7 @@
## Inspector Architecture
- Inspector source is `frontend/packages/inspector/`.
- `/ui/` must use ACP over HTTP (`/v2/rpc`) for session/prompt traffic.
- `/ui/` must use ACP over HTTP (`/v1/rpc`) for session/prompt traffic.
- Primary flow:
- `initialize`
- `session/new`

207
frontend/packages/inspector/index.html
View file

@ -110,6 +110,25 @@
color: var(--muted);
}
.header-link {
display: inline-flex;
align-items: center;
gap: 5px;
padding: 5px 10px;
border: 1px solid var(--border-2);
border-radius: var(--radius-sm);
font-size: 11px;
font-weight: 600;
color: var(--muted);
text-decoration: none;
transition: color var(--transition), border-color var(--transition);
}
.header-link:hover {
color: var(--accent);
border-color: var(--accent);
}
.status-indicator.disconnected {
display: flex;
align-items: center;
@ -404,7 +423,7 @@
flex-direction: column;
border-right: 1px solid var(--border);
background: var(--surface-2);
overflow: visible;
min-height: 0;
}
.sidebar-header {
@ -555,6 +574,20 @@
min-width: 0;
}
.setup-custom-back {
flex-shrink: 0;
background: none;
border: none;
color: var(--accent);
font-size: 10px;
cursor: pointer;
padding: 2px 4px;
}
.setup-custom-back:hover {
text-decoration: underline;
}
.session-create-section {
overflow: hidden;
}
@ -1055,6 +1088,23 @@
color: var(--danger);
}
.session-persistence-note {
padding: 8px 10px 10px;
border-top: 1px solid var(--border);
font-size: 10px;
line-height: 1.45;
color: var(--muted);
}
.session-persistence-note a {
color: var(--accent);
text-decoration: none;
}
.session-persistence-note a:hover {
text-decoration: underline;
}
/* Chat Panel */
.chat-panel {
display: flex;
@ -1322,6 +1372,64 @@
margin-top: 8px;
}
.toast-stack {
position: fixed;
right: 16px;
bottom: 16px;
z-index: 1000;
display: flex;
flex-direction: column;
gap: 8px;
width: min(420px, calc(100vw - 24px));
pointer-events: none;
}
.toast {
pointer-events: auto;
display: flex;
align-items: flex-start;
gap: 10px;
padding: 10px 12px;
border-radius: var(--radius-sm);
border: 1px solid rgba(255, 59, 48, 0.35);
background: rgba(28, 8, 8, 0.95);
box-shadow: 0 10px 25px rgba(0, 0, 0, 0.35);
}
.toast-content {
min-width: 0;
}
.toast-title {
font-size: 10px;
text-transform: uppercase;
letter-spacing: 0.4px;
color: var(--danger);
margin-bottom: 4px;
}
.toast-message {
color: var(--text-secondary);
font-size: 12px;
line-height: 1.4;
word-break: break-word;
}
.toast-close {
background: transparent;
border: none;
color: var(--muted);
font-size: 12px;
line-height: 1;
cursor: pointer;
padding: 2px;
margin-top: 1px;
}
.toast-close:hover {
color: var(--text);
}
.cursor {
display: inline-block;
width: 2px;
@ -1932,84 +2040,82 @@
letter-spacing: 0.3px;
}
.event-type.session,
.event-type.session-started,
.event-type.session-ended {
/* ACP event categories: connection */
.event-type.connection,
.event-type.session {
color: var(--success);
}
.event-type.item,
.event-type.item-started,
.event-type.item-completed {
color: var(--accent);
}
.event-type.item-delta {
color: var(--cyan);
}
.event-type.error,
.event-type.agent-unparsed {
color: var(--danger);
}
.event-type.question,
.event-type.question-requested,
.event-type.question-resolved {
color: var(--warning);
}
.event-type.permission,
.event-type.permission-requested,
.event-type.permission-resolved {
color: var(--purple);
}
.event-icon.session,
.event-icon.session-started,
.event-icon.session-ended {
.event-icon.connection,
.event-icon.session {
color: var(--success);
border-color: rgba(48, 209, 88, 0.35);
background: rgba(48, 209, 88, 0.12);
}
.event-icon.item,
.event-icon.item-started,
.event-icon.item-completed {
/* ACP event categories: prompt / tool */
.event-type.prompt,
.event-type.tool {
color: var(--accent);
}
.event-icon.prompt,
.event-icon.tool {
color: var(--accent);
border-color: rgba(255, 79, 0, 0.35);
background: rgba(255, 79, 0, 0.12);
}
.event-icon.item-delta {
/* ACP event categories: update / terminal (streaming, realtime) */
.event-type.update,
.event-type.terminal {
color: var(--cyan);
}
.event-icon.update,
.event-icon.terminal {
color: var(--cyan);
border-color: rgba(100, 210, 255, 0.35);
background: rgba(100, 210, 255, 0.12);
}
.event-icon.error,
.event-icon.agent-unparsed {
/* ACP event categories: cancel */
.event-type.cancel {
color: var(--danger);
}
.event-icon.cancel {
color: var(--danger);
border-color: rgba(255, 59, 48, 0.35);
background: rgba(255, 59, 48, 0.12);
}
.event-icon.question,
.event-icon.question-requested,
.event-icon.question-resolved {
/* ACP event categories: filesystem */
.event-type.filesystem {
color: var(--warning);
}
.event-icon.filesystem {
color: var(--warning);
border-color: rgba(255, 159, 10, 0.35);
background: rgba(255, 159, 10, 0.12);
}
.event-icon.permission,
.event-icon.permission-requested,
.event-icon.permission-resolved {
/* ACP event categories: config / permission */
.event-type.config,
.event-type.permission {
color: var(--purple);
}
.event-icon.config,
.event-icon.permission {
color: var(--purple);
border-color: rgba(191, 90, 242, 0.35);
background: rgba(191, 90, 242, 0.12);
}
/* ACP event categories: response (fallback) */
.event-type.response {
color: var(--muted);
}
.event-icon.response {
color: var(--muted);
}
.event-time {
font-size: 10px;
color: var(--muted);
@ -2247,6 +2353,13 @@
.header-title {
display: none;
}
.toast-stack {
left: 12px;
right: 12px;
bottom: 12px;
width: auto;
}
}
</style>
</head>

10
frontend/packages/inspector/package.json
View file

@ -6,19 +6,23 @@
"type": "module",
"scripts": {
"dev": "vite",
"build": "pnpm --filter sandbox-agent build && vite build",
"build": "SKIP_OPENAPI_GEN=1 pnpm --filter @sandbox-agent/persist-indexeddb build && vite build",
"preview": "vite preview",
"typecheck": "tsc --noEmit"
"typecheck": "SKIP_OPENAPI_GEN=1 pnpm --filter @sandbox-agent/persist-indexeddb build && tsc --noEmit",
"test": "SKIP_OPENAPI_GEN=1 pnpm --filter @sandbox-agent/persist-indexeddb build && vitest run"
},
"devDependencies": {
"sandbox-agent": "workspace:*",
"@types/react": "^18.3.3",
"@types/react-dom": "^18.3.0",
"@vitejs/plugin-react": "^4.3.1",
"fake-indexeddb": "^6.2.4",
"typescript": "^5.7.3",
"vite": "^5.4.7"
"vite": "^5.4.7",
"vitest": "^3.0.0"
},
"dependencies": {
"@sandbox-agent/persist-indexeddb": "workspace:*",
"lucide-react": "^0.469.0",
"react": "^18.3.1",
"react-dom": "^18.3.1"

29
frontend/packages/inspector/src/components/ConnectScreen.tsx
View file

@ -1,4 +1,4 @@
import { AlertTriangle, Zap } from "lucide-react";
import { AlertTriangle, BookOpen, Zap } from "lucide-react";
import { isHttpsToHttpConnection, isLocalNetworkTarget } from "../lib/permissions";
const logoUrl = `${import.meta.env.BASE_URL}logos/sandboxagent.svg`;
@ -11,7 +11,9 @@ const ConnectScreen = ({
onEndpointChange,
onTokenChange,
onConnect,
reportUrl
reportUrl,
docsUrl,
discordUrl,
}: {
endpoint: string;
token: string;
@ -21,6 +23,8 @@ const ConnectScreen = ({
onTokenChange: (value: string) => void;
onConnect: () => void;
reportUrl?: string;
docsUrl?: string;
discordUrl?: string;
}) => {
return (
<div className="app">
@ -28,11 +32,26 @@ const ConnectScreen = ({
<div className="header-left">
<img src={logoUrl} alt="Sandbox Agent" className="logo-text" style={{ height: '20px', width: 'auto' }} />
</div>
{reportUrl && (
{(docsUrl || discordUrl || reportUrl) && (
<div className="header-right">
<a className="button ghost small" href={reportUrl} target="_blank" rel="noreferrer">
Report Bug
{docsUrl && (
<a className="header-link" href={docsUrl} target="_blank" rel="noreferrer">
<BookOpen size={12} />
Docs
</a>
)}
{discordUrl && (
<a className="header-link" href={discordUrl} target="_blank" rel="noreferrer">
<svg width="12" height="12" viewBox="0 0 24 24" fill="currentColor"><path d="M20.317 4.37a19.791 19.791 0 0 0-4.885-1.515.074.074 0 0 0-.079.037c-.21.375-.444.864-.608 1.25a18.27 18.27 0 0 0-5.487 0 12.64 12.64 0 0 0-.617-1.25.077.077 0 0 0-.079-.037A19.736 19.736 0 0 0 3.677 4.37a.07.07 0 0 0-.032.027C.533 9.046-.32 13.58.099 18.057a.082.082 0 0 0 .031.057 19.9 19.9 0 0 0 5.993 3.03.078.078 0 0 0 .084-.028c.462-.63.874-1.295 1.226-1.994a.076.076 0 0 0-.041-.106 13.107 13.107 0 0 1-1.872-.892.077.077 0 0 1-.008-.128 10.2 10.2 0 0 0 .372-.292.074.074 0 0 1 .077-.01c3.928 1.793 8.18 1.793 12.062 0a.074.074 0 0 1 .078.01c.12.098.246.198.373.292a.077.077 0 0 1-.006.127 12.299 12.299 0 0 1-1.873.892.077.077 0 0 0-.041.107c.36.698.772 1.362 1.225 1.993a.076.076 0 0 0 .084.028 19.839 19.839 0 0 0 6.002-3.03.077.077 0 0 0 .032-.054c.5-5.177-.838-9.674-3.549-13.66a.061.061 0 0 0-.031-.03zM8.02 15.33c-1.183 0-2.157-1.085-2.157-2.419 0-1.333.956-2.419 2.157-2.419 1.21 0 2.176 1.095 2.157 2.42 0 1.333-.956 2.418-2.157 2.418zm7.975 0c-1.183 0-2.157-1.085-2.157-2.419 0-1.333.955-2.419 2.157-2.419 1.21 0 2.176 1.095 2.157 2.42 0 1.333-.946 2.418-2.157 2.418z"/></svg>
Discord
</a>
)}
{reportUrl && (
<a className="header-link" href={reportUrl} target="_blank" rel="noreferrer">
<svg width="12" height="12" viewBox="0 0 24 24" fill="currentColor"><path d="M12 .297c-6.63 0-12 5.373-12 12 0 5.303 3.438 9.8 8.205 11.385.6.113.82-.258.82-.577 0-.285-.01-1.04-.015-2.04-3.338.724-4.042-1.61-4.042-1.61C4.422 18.07 3.633 17.7 3.633 17.7c-1.087-.744.084-.729.084-.729 1.205.084 1.838 1.236 1.838 1.236 1.07 1.835 2.809 1.305 3.495.998.108-.776.417-1.305.76-1.605-2.665-.3-5.466-1.332-5.466-5.93 0-1.31.465-2.38 1.235-3.22-.135-.303-.54-1.523.105-3.176 0 0 1.005-.322 3.3 1.23.96-.267 1.98-.399 3-.405 1.02.006 2.04.138 3 .405 2.28-1.552 3.285-1.23 3.285-1.23.645 1.653.24 2.873.12 3.176.765.84 1.23 1.91 1.23 3.22 0 4.61-2.805 5.625-5.475 5.92.42.36.81 1.096.81 2.22 0 1.606-.015 2.896-.015 3.286 0 .315.21.69.825.57C20.565 22.092 24 17.592 24 12.297c0-6.627-5.373-12-12-12"/></svg>
Issues
</a>
)}
</div>
)}
</header>

719
frontend/packages/inspector/src/components/SessionCreateMenu.tsx
View file

@ -1,15 +1,17 @@
import { ArrowLeft, ArrowRight, ChevronDown, ChevronRight, Pencil, Plus, X } from "lucide-react";
import { useEffect, useRef, useState } from "react";
import type { McpServerEntry } from "../App";
import type { AgentInfo, AgentModelInfo, AgentModeInfo, SkillSource } from "../types/legacyApi";
import { ArrowLeft, ArrowRight } from "lucide-react";
import { useEffect, useState } from "react";
import type { AgentInfo } from "sandbox-agent";
type AgentModeInfo = { id: string; name: string; description: string };
type AgentModelInfo = { id: string; name?: string };
export type SessionConfig = {
model: string;
agentMode: string;
permissionMode: string;
variant: string;
model: string;
};
const CUSTOM_MODEL_VALUE = "__custom__";
const agentLabels: Record<string, string> = {
claude: "Claude Code",
codex: "Codex",
@ -17,59 +19,6 @@ const agentLabels: Record<string, string> = {
amp: "Amp"
};
const validateServerJson = (json: string): string | null => {
const trimmed = json.trim();
if (!trimmed) return "Config is required";
try {
const parsed = JSON.parse(trimmed);
if (parsed === null || typeof parsed !== "object" || Array.isArray(parsed)) {
return "Must be a JSON object";
}
if (!parsed.type) return 'Missing "type" field';
if (parsed.type !== "local" && parsed.type !== "remote") {
return 'Type must be "local" or "remote"';
}
if (parsed.type === "local" && !parsed.command) return 'Local server requires "command"';
if (parsed.type === "remote" && !parsed.url) return 'Remote server requires "url"';
return null;
} catch {
return "Invalid JSON";
}
};
const getServerType = (configJson: string): string | null => {
try {
const parsed = JSON.parse(configJson);
return parsed?.type ?? null;
} catch {
return null;
}
};
const getServerSummary = (configJson: string): string => {
try {
const parsed = JSON.parse(configJson);
if (parsed?.type === "local") {
const cmd = Array.isArray(parsed.command) ? parsed.command.join(" ") : parsed.command;
return cmd ?? "local";
}
if (parsed?.type === "remote") {
return parsed.url ?? "remote";
}
return parsed?.type ?? "";
} catch {
return "";
}
};
const skillSourceSummary = (source: SkillSource): string => {
let summary = source.source;
if (source.skills && source.skills.length > 0) {
summary += ` [${source.skills.join(", ")}]`;
}
return summary;
};
const SessionCreateMenu = ({
agents,
agentsLoading,
@ -77,17 +26,8 @@ const SessionCreateMenu = ({
modesByAgent,
modelsByAgent,
defaultModelByAgent,
modesLoadingByAgent,
modelsLoadingByAgent,
modesErrorByAgent,
modelsErrorByAgent,
mcpServers,
onMcpServersChange,
mcpConfigError,
skillSources,
onSkillSourcesChange,
onSelectAgent,
onCreateSession,
onSelectAgent,
open,
onClose
}: {
@ -97,60 +37,18 @@ const SessionCreateMenu = ({
modesByAgent: Record<string, AgentModeInfo[]>;
modelsByAgent: Record<string, AgentModelInfo[]>;
defaultModelByAgent: Record<string, string>;
modesLoadingByAgent: Record<string, boolean>;
modelsLoadingByAgent: Record<string, boolean>;
modesErrorByAgent: Record<string, string | null>;
modelsErrorByAgent: Record<string, string | null>;
mcpServers: McpServerEntry[];
onMcpServersChange: (servers: McpServerEntry[]) => void;
mcpConfigError: string | null;
skillSources: SkillSource[];
onSkillSourcesChange: (sources: SkillSource[]) => void;
onSelectAgent: (agentId: string) => void;
onCreateSession: (agentId: string, config: SessionConfig) => void;
onSelectAgent: (agentId: string) => Promise<void>;
open: boolean;
onClose: () => void;
}) => {
const [phase, setPhase] = useState<"agent" | "config">("agent");
const [phase, setPhase] = useState<"agent" | "config" | "loading-config">("agent");
const [selectedAgent, setSelectedAgent] = useState("");
const [agentMode, setAgentMode] = useState("");
const [permissionMode, setPermissionMode] = useState("default");
const [model, setModel] = useState("");
const [variant, setVariant] = useState("");
const [mcpExpanded, setMcpExpanded] = useState(false);
const [skillsExpanded, setSkillsExpanded] = useState(false);
// Skill add/edit state
const [addingSkill, setAddingSkill] = useState(false);
const [editingSkillIndex, setEditingSkillIndex] = useState<number | null>(null);
const [skillType, setSkillType] = useState<"github" | "local" | "git">("github");
const [skillSource, setSkillSource] = useState("");
const [skillFilter, setSkillFilter] = useState("");
const [skillRef, setSkillRef] = useState("");
const [skillSubpath, setSkillSubpath] = useState("");
const [skillLocalError, setSkillLocalError] = useState<string | null>(null);
const skillSourceRef = useRef<HTMLInputElement>(null);
// MCP add/edit state
const [addingMcp, setAddingMcp] = useState(false);
const [editingMcpIndex, setEditingMcpIndex] = useState<number | null>(null);
const [mcpName, setMcpName] = useState("");
const [mcpJson, setMcpJson] = useState("");
const [mcpLocalError, setMcpLocalError] = useState<string | null>(null);
const mcpNameRef = useRef<HTMLInputElement>(null);
const mcpJsonRef = useRef<HTMLTextAreaElement>(null);
const cancelSkillEdit = () => {
setAddingSkill(false);
setEditingSkillIndex(null);
setSkillType("github");
setSkillSource("");
setSkillFilter("");
setSkillRef("");
setSkillSubpath("");
setSkillLocalError(null);
};
const [selectedModel, setSelectedModel] = useState("");
const [customModel, setCustomModel] = useState("");
const [isCustomModel, setIsCustomModel] = useState(false);
const [configLoadDone, setConfigLoadDone] = useState(false);
// Reset state when menu closes
useEffect(() => {
@ -158,20 +56,21 @@ const SessionCreateMenu = ({
setPhase("agent");
setSelectedAgent("");
setAgentMode("");
setPermissionMode("default");
setModel("");
setVariant("");
setMcpExpanded(false);
setSkillsExpanded(false);
cancelSkillEdit();
setAddingMcp(false);
setEditingMcpIndex(null);
setMcpName("");
setMcpJson("");
setMcpLocalError(null);
setSelectedModel("");
setCustomModel("");
setIsCustomModel(false);
setConfigLoadDone(false);
}
}, [open]);
// Transition to config phase after load completes — deferred via useEffect
// so parent props (modelsByAgent) have settled before we render the config form
useEffect(() => {
if (phase === "loading-config" && configLoadDone) {
setPhase("config");
}
}, [phase, configLoadDone]);
// Auto-select first mode when modes load for selected agent
useEffect(() => {
if (!selectedAgent) return;
@ -181,174 +80,60 @@ const SessionCreateMenu = ({
}
}, [modesByAgent, selectedAgent, agentMode]);
// Focus skill source input when adding
// Auto-select default model when agent is selected
useEffect(() => {
if ((addingSkill || editingSkillIndex !== null) && skillSourceRef.current) {
skillSourceRef.current.focus();
if (!selectedAgent) return;
if (selectedModel) return;
const defaultModel = defaultModelByAgent[selectedAgent];
if (defaultModel) {
setSelectedModel(defaultModel);
} else {
const models = modelsByAgent[selectedAgent];
if (models && models.length > 0) {
setSelectedModel(models[0].id);
}
}, [addingSkill, editingSkillIndex]);
// Focus MCP name input when adding
useEffect(() => {
if (addingMcp && mcpNameRef.current) {
mcpNameRef.current.focus();
}
}, [addingMcp]);
// Focus MCP json textarea when editing
useEffect(() => {
if (editingMcpIndex !== null && mcpJsonRef.current) {
mcpJsonRef.current.focus();
}
}, [editingMcpIndex]);
}, [modelsByAgent, defaultModelByAgent, selectedAgent, selectedModel]);
if (!open) return null;
const handleAgentClick = (agentId: string) => {
setSelectedAgent(agentId);
setPhase("config");
onSelectAgent(agentId);
setPhase("loading-config");
setConfigLoadDone(false);
onSelectAgent(agentId).finally(() => {
setConfigLoadDone(true);
});
};
const handleBack = () => {
setPhase("agent");
setSelectedAgent("");
setAgentMode("");
setPermissionMode("default");
setModel("");
setVariant("");
setSelectedModel("");
setCustomModel("");
setIsCustomModel(false);
setConfigLoadDone(false);
};
const handleModelSelectChange = (value: string) => {
if (value === CUSTOM_MODEL_VALUE) {
setIsCustomModel(true);
setSelectedModel("");
} else {
setIsCustomModel(false);
setCustomModel("");
setSelectedModel(value);
}
};
const resolvedModel = isCustomModel ? customModel : selectedModel;
const handleCreate = () => {
if (mcpConfigError) return;
onCreateSession(selectedAgent, { model, agentMode, permissionMode, variant });
onCreateSession(selectedAgent, { agentMode, model: resolvedModel });
onClose();
};
// Skill source helpers
const startAddSkill = () => {
setAddingSkill(true);
setEditingSkillIndex(null);
setSkillType("github");
setSkillSource("rivet-dev/skills");
setSkillFilter("sandbox-agent");
setSkillRef("");
setSkillSubpath("");
setSkillLocalError(null);
};
const startEditSkill = (index: number) => {
const entry = skillSources[index];
setEditingSkillIndex(index);
setAddingSkill(false);
setSkillType(entry.type as "github" | "local" | "git");
setSkillSource(entry.source);
setSkillFilter(entry.skills?.join(", ") ?? "");
setSkillRef(entry.ref ?? "");
setSkillSubpath(entry.subpath ?? "");
setSkillLocalError(null);
};
const commitSkill = () => {
const src = skillSource.trim();
if (!src) {
setSkillLocalError("Source is required");
return;
}
const entry: SkillSource = {
type: skillType,
source: src,
};
const filterList = skillFilter.trim()
? skillFilter.split(",").map((s) => s.trim()).filter(Boolean)
: undefined;
if (filterList && filterList.length > 0) entry.skills = filterList;
if (skillRef.trim()) entry.ref = skillRef.trim();
if (skillSubpath.trim()) entry.subpath = skillSubpath.trim();
if (editingSkillIndex !== null) {
const updated = [...skillSources];
updated[editingSkillIndex] = entry;
onSkillSourcesChange(updated);
} else {
onSkillSourcesChange([...skillSources, entry]);
}
cancelSkillEdit();
};
const removeSkill = (index: number) => {
onSkillSourcesChange(skillSources.filter((_, i) => i !== index));
if (editingSkillIndex === index) {
cancelSkillEdit();
}
};
const isEditingSkill = addingSkill || editingSkillIndex !== null;
const startAddMcp = () => {
setAddingMcp(true);
setEditingMcpIndex(null);
setMcpName("everything");
setMcpJson('{\n "type": "local",\n "command": "npx",\n "args": ["@modelcontextprotocol/server-everything"]\n}');
setMcpLocalError(null);
};
const startEditMcp = (index: number) => {
const entry = mcpServers[index];
setEditingMcpIndex(index);
setAddingMcp(false);
setMcpName(entry.name);
setMcpJson(entry.configJson);
setMcpLocalError(entry.error);
};
const cancelMcpEdit = () => {
setAddingMcp(false);
setEditingMcpIndex(null);
setMcpName("");
setMcpJson("");
setMcpLocalError(null);
};
const commitMcp = () => {
const name = mcpName.trim();
if (!name) {
setMcpLocalError("Server name is required");
return;
}
const error = validateServerJson(mcpJson);
if (error) {
setMcpLocalError(error);
return;
}
// Check for duplicate names (except when editing the same entry)
const duplicate = mcpServers.findIndex((e) => e.name === name);
if (duplicate !== -1 && duplicate !== editingMcpIndex) {
setMcpLocalError(`Server "${name}" already exists`);
return;
}
const entry: McpServerEntry = { name, configJson: mcpJson.trim(), error: null };
if (editingMcpIndex !== null) {
const updated = [...mcpServers];
updated[editingMcpIndex] = entry;
onMcpServersChange(updated);
} else {
onMcpServersChange([...mcpServers, entry]);
}
cancelMcpEdit();
};
const removeMcp = (index: number) => {
onMcpServersChange(mcpServers.filter((_, i) => i !== index));
if (editingMcpIndex === index) {
cancelMcpEdit();
}
};
const isEditingMcp = addingMcp || editingMcpIndex !== null;
if (phase === "agent") {
return (
<div className="session-create-menu">
@ -378,30 +163,25 @@ const SessionCreateMenu = ({
);
}
const agentLabel = agentLabels[selectedAgent] ?? selectedAgent;
if (phase === "loading-config") {
return (
<div className="session-create-menu">
<div className="session-create-header">
<button className="session-create-back" onClick={handleBack} title="Back to agents">
<ArrowLeft size={14} />
</button>
<span className="session-create-agent-name">{agentLabel}</span>
</div>
<div className="sidebar-add-status">Loading config...</div>
</div>
);
}
// Phase 2: config form
const activeModes = modesByAgent[selectedAgent] ?? [];
const modesLoading = modesLoadingByAgent[selectedAgent] ?? false;
const modesError = modesErrorByAgent[selectedAgent] ?? null;
const modelOptions = modelsByAgent[selectedAgent] ?? [];
const modelsLoading = modelsLoadingByAgent[selectedAgent] ?? false;
const modelsError = modelsErrorByAgent[selectedAgent] ?? null;
const defaultModel = defaultModelByAgent[selectedAgent] ?? "";
const selectedModelId = model || defaultModel;
const selectedModelObj = modelOptions.find((entry) => entry.id === selectedModelId);
const variantOptions = selectedModelObj?.variants ?? [];
const showModelSelect = modelsLoading || Boolean(modelsError) || modelOptions.length > 0;
const hasModelOptions = modelOptions.length > 0;
const modelCustom =
model && hasModelOptions && !modelOptions.some((entry) => entry.id === model);
const supportsVariants =
modelsLoading ||
Boolean(modelsError) ||
modelOptions.some((entry) => (entry.variants?.length ?? 0) > 0);
const showVariantSelect =
supportsVariants && (modelsLoading || Boolean(modelsError) || variantOptions.length > 0);
const hasVariantOptions = variantOptions.length > 0;
const variantCustom = variant && hasVariantOptions && !variantOptions.includes(variant);
const agentLabel = agentLabels[selectedAgent] ?? selectedAgent;
const activeModels = modelsByAgent[selectedAgent] ?? [];
return (
<div className="session-create-menu">
@ -415,43 +195,49 @@ const SessionCreateMenu = ({
<div className="session-create-form">
<div className="setup-field">
<span className="setup-label">Model</span>
{showModelSelect ? (
<select
className="setup-select"
value={model}
onChange={(e) => { setModel(e.target.value); setVariant(""); }}
title="Model"
disabled={modelsLoading || Boolean(modelsError)}
>
{modelsLoading ? (
<option value="">Loading models...</option>
) : modelsError ? (
<option value="">{modelsError}</option>
) : (
<>
<option value="">
{defaultModel ? `Default (${defaultModel})` : "Default"}
</option>
{modelCustom && <option value={model}>{model} (custom)</option>}
{modelOptions.map((entry) => (
<option key={entry.id} value={entry.id}>
{entry.name ?? entry.id}
</option>
))}
</>
)}
</select>
) : (
{isCustomModel ? (
<input
className="setup-input"
value={model}
onChange={(e) => setModel(e.target.value)}
placeholder="Model"
title="Model"
type="text"
value={customModel}
onChange={(e) => setCustomModel(e.target.value)}
placeholder="Enter model name..."
autoFocus
/>
) : (
<select
className="setup-select"
value={selectedModel}
onChange={(e) => handleModelSelectChange(e.target.value)}
title="Model"
>
{activeModels.map((m) => (
<option key={m.id} value={m.id}>
{m.name || m.id}
</option>
))}
<option value={CUSTOM_MODEL_VALUE}>Custom...</option>
</select>
)}
{isCustomModel && (
<button
className="setup-custom-back"
onClick={() => {
setIsCustomModel(false);
setCustomModel("");
const defaultModel = defaultModelByAgent[selectedAgent];
setSelectedModel(
defaultModel || (activeModels.length > 0 ? activeModels[0].id : "")
);
}}
title="Back to model list"
type="button"
>
← List
</button>
)}
</div>
{activeModes.length > 0 && (
<div className="setup-field">
<span className="setup-label">Mode</span>
<select
@ -459,286 +245,19 @@ const SessionCreateMenu = ({
value={agentMode}
onChange={(e) => setAgentMode(e.target.value)}
title="Mode"
disabled={modesLoading || Boolean(modesError)}
>
{modesLoading ? (
<option value="">Loading modes...</option>
) : modesError ? (
<option value="">{modesError}</option>
) : activeModes.length > 0 ? (
activeModes.map((m) => (
{activeModes.map((m) => (
<option key={m.id} value={m.id}>
{m.name || m.id}
</option>
))
) : (
<option value="">Mode</option>
)}
</select>
</div>
<div className="setup-field">
<span className="setup-label">Permission</span>
<select
className="setup-select"
value={permissionMode}
onChange={(e) => setPermissionMode(e.target.value)}
title="Permission Mode"
>
<option value="default">Default</option>
<option value="plan">Plan</option>
<option value="bypass">Bypass</option>
</select>
</div>
{supportsVariants && (
<div className="setup-field">
<span className="setup-label">Variant</span>
{showVariantSelect ? (
<select
className="setup-select"
value={variant}
onChange={(e) => setVariant(e.target.value)}
title="Variant"
disabled={modelsLoading || Boolean(modelsError)}
>
{modelsLoading ? (
<option value="">Loading variants...</option>
) : modelsError ? (
<option value="">{modelsError}</option>
) : (
<>
<option value="">Default</option>
{variantCustom && <option value={variant}>{variant} (custom)</option>}
{variantOptions.map((entry) => (
<option key={entry} value={entry}>
{entry}
</option>
))}
</>
)}
</select>
) : (
<input
className="setup-input"
value={variant}
onChange={(e) => setVariant(e.target.value)}
placeholder="Variant"
title="Variant"
/>
)}
</div>
)}
{/* MCP Servers - collapsible */}
<div className="session-create-section">
<button
type="button"
className="session-create-section-toggle"
onClick={() => setMcpExpanded(!mcpExpanded)}
>
<span className="setup-label">MCP</span>
<span className="session-create-section-count">{mcpServers.length} server{mcpServers.length !== 1 ? "s" : ""}</span>
{mcpExpanded ? <ChevronDown size={12} className="session-create-section-arrow" /> : <ChevronRight size={12} className="session-create-section-arrow" />}
</button>
{mcpExpanded && (
<div className="session-create-section-body">
{mcpServers.length > 0 && !isEditingMcp && (
<div className="session-create-mcp-list">
{mcpServers.map((entry, index) => (
<div key={entry.name} className="session-create-mcp-item">
<div className="session-create-mcp-info">
<span className="session-create-mcp-name">{entry.name}</span>
{getServerType(entry.configJson) && (
<span className="session-create-mcp-type">{getServerType(entry.configJson)}</span>
)}
<span className="session-create-mcp-summary mono">{getServerSummary(entry.configJson)}</span>
</div>
<div className="session-create-mcp-actions">
<button
type="button"
className="session-create-skill-remove"
onClick={() => startEditMcp(index)}
title="Edit server"
>
<Pencil size={10} />
</button>
<button
type="button"
className="session-create-skill-remove"
onClick={() => removeMcp(index)}
title="Remove server"
>
<X size={12} />
</button>
</div>
</div>
))}
</div>
)}
{isEditingMcp ? (
<div className="session-create-mcp-edit">
<input
ref={mcpNameRef}
className="session-create-mcp-name-input"
value={mcpName}
onChange={(e) => { setMcpName(e.target.value); setMcpLocalError(null); }}
placeholder="server-name"
disabled={editingMcpIndex !== null}
/>
<textarea
ref={mcpJsonRef}
className="session-create-textarea mono"
value={mcpJson}
onChange={(e) => { setMcpJson(e.target.value); setMcpLocalError(null); }}
placeholder='{"type":"local","command":"node","args":["./server.js"]}'
rows={4}
/>
{mcpLocalError && (
<div className="session-create-inline-error">{mcpLocalError}</div>
)}
<div className="session-create-mcp-edit-actions">
<button type="button" className="session-create-mcp-save" onClick={commitMcp}>
{editingMcpIndex !== null ? "Save" : "Add"}
</button>
<button type="button" className="session-create-mcp-cancel" onClick={cancelMcpEdit}>
Cancel
</button>
</div>
</div>
) : (
<button
type="button"
className="session-create-add-btn"
onClick={startAddMcp}
>
<Plus size={12} />
Add server
</button>
)}
{mcpConfigError && !isEditingMcp && (
<div className="session-create-inline-error">{mcpConfigError}</div>
)}
</div>
)}
</div>
{/* Skills - collapsible with source-based list */}
<div className="session-create-section">
<button
type="button"
className="session-create-section-toggle"
onClick={() => setSkillsExpanded(!skillsExpanded)}
>
<span className="setup-label">Skills</span>
<span className="session-create-section-count">{skillSources.length} source{skillSources.length !== 1 ? "s" : ""}</span>
{skillsExpanded ? <ChevronDown size={12} className="session-create-section-arrow" /> : <ChevronRight size={12} className="session-create-section-arrow" />}
</button>
{skillsExpanded && (
<div className="session-create-section-body">
{skillSources.length > 0 && !isEditingSkill && (
<div className="session-create-skill-list">
{skillSources.map((entry, index) => (
<div key={`${entry.type}-${entry.source}-${index}`} className="session-create-skill-item">
<span className="session-create-skill-type-badge">{entry.type}</span>
<span className="session-create-skill-path mono">{skillSourceSummary(entry)}</span>
<div className="session-create-mcp-actions">
<button
type="button"
className="session-create-skill-remove"
onClick={() => startEditSkill(index)}
title="Edit source"
>
<Pencil size={10} />
</button>
<button
type="button"
className="session-create-skill-remove"
onClick={() => removeSkill(index)}
title="Remove source"
>
<X size={12} />
</button>
</div>
</div>
))}
</div>
)}
{isEditingSkill ? (
<div className="session-create-mcp-edit">
<div className="session-create-skill-type-row">
<select
className="session-create-skill-type-select"
value={skillType}
onChange={(e) => { setSkillType(e.target.value as "github" | "local" | "git"); setSkillLocalError(null); }}
>
<option value="github">github</option>
<option value="local">local</option>
<option value="git">git</option>
</select>
<input
ref={skillSourceRef}
className="session-create-skill-input mono"
value={skillSource}
onChange={(e) => { setSkillSource(e.target.value); setSkillLocalError(null); }}
placeholder={skillType === "github" ? "owner/repo" : skillType === "local" ? "/path/to/skill" : "https://git.example.com/repo.git"}
/>
</div>
<input
className="session-create-skill-input mono"
value={skillFilter}
onChange={(e) => setSkillFilter(e.target.value)}
placeholder="Filter skills (comma-separated, optional)"
/>
{skillType !== "local" && (
<div className="session-create-skill-type-row">
<input
className="session-create-skill-input mono"
value={skillRef}
onChange={(e) => setSkillRef(e.target.value)}
placeholder="Branch/tag (optional)"
/>
<input
className="session-create-skill-input mono"
value={skillSubpath}
onChange={(e) => setSkillSubpath(e.target.value)}
placeholder="Subpath (optional)"
/>
</div>
)}
{skillLocalError && (
<div className="session-create-inline-error">{skillLocalError}</div>
)}
<div className="session-create-mcp-edit-actions">
<button type="button" className="session-create-mcp-save" onClick={commitSkill}>
{editingSkillIndex !== null ? "Save" : "Add"}
</button>
<button type="button" className="session-create-mcp-cancel" onClick={cancelSkillEdit}>
Cancel
</button>
</div>
</div>
) : (
<button
type="button"
className="session-create-add-btn"
onClick={startAddSkill}
>
<Plus size={12} />
Add source
</button>
)}
</div>
)}
</div>
</div>
<div className="session-create-actions">
<button
className="button primary"
onClick={handleCreate}
disabled={Boolean(mcpConfigError)}
>
<button className="button primary" onClick={handleCreate}>
Create Session
</button>
</div>

83
frontend/packages/inspector/src/components/chat/ChatMessages.tsx
View file

@ -1,23 +1,22 @@
import { getAvatarLabel, getMessageClass } from "./messageUtils";
import renderContentPart from "./renderContentPart";
import type { TimelineEntry } from "./types";
import { formatJson } from "../../utils/format";
const ChatMessages = ({
entries,
sessionError,
eventError,
messagesEndRef
}: {
entries: TimelineEntry[];
sessionError: string | null;
eventError: string | null;
messagesEndRef: React.RefObject<HTMLDivElement>;
}) => {
return (
<div className="messages">
{entries.map((entry) => {
const messageClass = getMessageClass(entry);
if (entry.kind === "meta") {
const messageClass = entry.meta?.severity === "error" ? "error" : "system";
return (
<div key={entry.id} className={`message ${messageClass}`}>
<div className="avatar">{getAvatarLabel(messageClass)}</div>
@ -31,53 +30,73 @@ const ChatMessages = ({
);
}
const item = entry.item;
if (!item) return null;
const hasParts = (item.content ?? []).length > 0;
const isInProgress = item.status === "in_progress";
const isFailed = item.status === "failed";
const messageClass = getMessageClass(item);
const statusValue = item.status ?? "";
const statusLabel =
statusValue && statusValue !== "completed" ? statusValue.replace("_", " ") : "";
const kindLabel = item.kind.replace("_", " ");
if (entry.kind === "reasoning") {
return (
<div key={entry.id} className={`message ${messageClass} ${isFailed ? "error" : ""}`}>
<div className="avatar">{getAvatarLabel(isFailed ? "error" : messageClass)}</div>
<div key={entry.id} className="message assistant">
<div className="avatar">AI</div>
<div className="message-content">
{(item.kind !== "message" || item.status !== "completed") && (
<div className="message-meta">
<span>{kindLabel}</span>
{statusLabel && (
<span className={`pill ${item.status === "failed" ? "danger" : "accent"}`}>
{statusLabel}
<span>reasoning - {entry.reasoning?.visibility ?? "public"}</span>
</div>
<div className="part-body muted">{entry.reasoning?.text ?? ""}</div>
</div>
</div>
);
}
if (entry.kind === "tool") {
const isComplete = entry.toolStatus === "completed" || entry.toolStatus === "failed";
const isFailed = entry.toolStatus === "failed";
return (
<div key={entry.id} className={`message tool ${isFailed ? "error" : ""}`}>
<div className="avatar">{getAvatarLabel(isFailed ? "error" : "tool")}</div>
<div className="message-content">
<div className="message-meta">
<span>tool call - {entry.toolName}</span>
{entry.toolStatus && entry.toolStatus !== "completed" && (
<span className={`pill ${isFailed ? "danger" : "accent"}`}>
{entry.toolStatus.replace("_", " ")}
</span>
)}
</div>
{entry.toolInput && <pre className="code-block">{entry.toolInput}</pre>}
{isComplete && entry.toolOutput && (
<div className="part">
<div className="part-title">result</div>
<pre className="code-block">{entry.toolOutput}</pre>
</div>
)}
{hasParts ? (
(item.content ?? []).map(renderContentPart)
) : entry.deltaText ? (
<span>
{entry.deltaText}
{isInProgress && <span className="cursor" />}
</span>
) : isInProgress ? (
{!isComplete && !entry.toolInput && (
<span className="thinking-indicator">
<span className="thinking-dot" />
<span className="thinking-dot" />
<span className="thinking-dot" />
</span>
)}
</div>
</div>
);
}
// Message (user or assistant)
return (
<div key={entry.id} className={`message ${messageClass}`}>
<div className="avatar">{getAvatarLabel(messageClass)}</div>
<div className="message-content">
{entry.text ? (
<div className="part-body">{entry.text}</div>
) : (
<span className="muted">No content yet.</span>
<span className="thinking-indicator">
<span className="thinking-dot" />
<span className="thinking-dot" />
<span className="thinking-dot" />
</span>
)}
</div>
</div>
);
})}
{sessionError && <div className="message-error">{sessionError}</div>}
{eventError && <div className="message-error">{eventError}</div>}
<div ref={messagesEndRef} />
</div>
);

122
frontend/packages/inspector/src/components/chat/ChatPanel.tsx
View file

@ -1,15 +1,9 @@
import { MessageSquare, Plus, Square, Terminal } from "lucide-react";
import { CheckSquare, MessageSquare, Plus, Square, Terminal } from "lucide-react";
import { useEffect, useRef, useState } from "react";
import type { McpServerEntry } from "../../App";
import type {
AgentInfo,
AgentModelInfo,
AgentModeInfo,
PermissionEventData,
QuestionEventData,
SkillSource
} from "../../types/legacyApi";
import ApprovalsTab from "../debug/ApprovalsTab";
import type { AgentInfo } from "sandbox-agent";
type AgentModeInfo = { id: string; name: string; description: string };
type AgentModelInfo = { id: string; name?: string };
import SessionCreateMenu, { type SessionConfig } from "../SessionCreateMenu";
import ChatInput from "./ChatInput";
import ChatMessages from "./ChatMessages";
@ -31,32 +25,11 @@ const ChatPanel = ({
messagesEndRef,
agentLabel,
currentAgentVersion,
sessionModel,
sessionVariant,
sessionPermissionMode,
sessionMcpServerCount,
sessionSkillSourceCount,
sessionEnded,
onEndSession,
eventError,
questionRequests,
permissionRequests,
questionSelections,
onSelectQuestionOption,
onAnswerQuestion,
onRejectQuestion,
onReplyPermission,
modesByAgent,
modelsByAgent,
defaultModelByAgent,
modesLoadingByAgent,
modelsLoadingByAgent,
modesErrorByAgent,
modelsErrorByAgent,
mcpServers,
onMcpServersChange,
mcpConfigError,
skillSources,
onSkillSourcesChange
}: {
sessionId: string;
transcriptEntries: TimelineEntry[];
@ -66,39 +39,18 @@ const ChatPanel = ({
onSendMessage: () => void;
onKeyDown: (event: React.KeyboardEvent<HTMLTextAreaElement>) => void;
onCreateSession: (agentId: string, config: SessionConfig) => void;
onSelectAgent: (agentId: string) => void;
onSelectAgent: (agentId: string) => Promise<void>;
agents: AgentInfo[];
agentsLoading: boolean;
agentsError: string | null;
messagesEndRef: React.RefObject<HTMLDivElement>;
agentLabel: string;
currentAgentVersion?: string | null;
sessionModel?: string | null;
sessionVariant?: string | null;
sessionPermissionMode?: string | null;
sessionMcpServerCount: number;
sessionSkillSourceCount: number;
sessionEnded: boolean;
onEndSession: () => void;
eventError: string | null;
questionRequests: QuestionEventData[];
permissionRequests: PermissionEventData[];
questionSelections: Record<string, string[][]>;
onSelectQuestionOption: (requestId: string, optionLabel: string) => void;
onAnswerQuestion: (request: QuestionEventData) => void;
onRejectQuestion: (requestId: string) => void;
onReplyPermission: (requestId: string, reply: "once" | "always" | "reject") => void;
modesByAgent: Record<string, AgentModeInfo[]>;
modelsByAgent: Record<string, AgentModelInfo[]>;
defaultModelByAgent: Record<string, string>;
modesLoadingByAgent: Record<string, boolean>;
modelsLoadingByAgent: Record<string, boolean>;
modesErrorByAgent: Record<string, string | null>;
modelsErrorByAgent: Record<string, string | null>;
mcpServers: McpServerEntry[];
onMcpServersChange: (servers: McpServerEntry[]) => void;
mcpConfigError: string | null;
skillSources: SkillSource[];
onSkillSourcesChange: (sources: SkillSource[]) => void;
}) => {
const [showAgentMenu, setShowAgentMenu] = useState(false);
const menuRef = useRef<HTMLDivElement | null>(null);
@ -115,8 +67,6 @@ const ChatPanel = ({
return () => document.removeEventListener("mousedown", handler);
}, [showAgentMenu]);
const hasApprovals = questionRequests.length > 0 || permissionRequests.length > 0;
return (
<div className="chat-panel">
<div className="panel-header">
@ -127,6 +77,12 @@ const ChatPanel = ({
</div>
<div className="panel-header-right">
{sessionId && (
sessionEnded ? (
<span className="button ghost small" style={{ opacity: 0.5, cursor: "default" }} title="Session ended">
<CheckSquare size={12} />
Ended
</span>
) : (
<button
type="button"
className="button ghost small"
@ -136,6 +92,7 @@ const ChatPanel = ({
<Square size={12} />
End
</button>
)
)}
</div>
</div>
@ -161,17 +118,8 @@ const ChatPanel = ({
modesByAgent={modesByAgent}
modelsByAgent={modelsByAgent}
defaultModelByAgent={defaultModelByAgent}
modesLoadingByAgent={modesLoadingByAgent}
modelsLoadingByAgent={modelsLoadingByAgent}
modesErrorByAgent={modesErrorByAgent}
modelsErrorByAgent={modelsErrorByAgent}
mcpServers={mcpServers}
onMcpServersChange={onMcpServersChange}
mcpConfigError={mcpConfigError}
skillSources={skillSources}
onSkillSourcesChange={onSkillSourcesChange}
onSelectAgent={onSelectAgent}
onCreateSession={onCreateSession}
onSelectAgent={onSelectAgent}
open={showAgentMenu}
onClose={() => setShowAgentMenu(false)}
/>
@ -187,27 +135,11 @@ const ChatPanel = ({
<ChatMessages
entries={transcriptEntries}
sessionError={sessionError}
eventError={eventError}
messagesEndRef={messagesEndRef}
/>
)}
</div>
{hasApprovals && (
<div className="approvals-inline">
<div className="approvals-inline-header">Approvals</div>
<ApprovalsTab
questionRequests={questionRequests}
permissionRequests={permissionRequests}
questionSelections={questionSelections}
onSelectQuestionOption={onSelectQuestionOption}
onAnswerQuestion={onAnswerQuestion}
onRejectQuestion={onRejectQuestion}
onReplyPermission={onReplyPermission}
/>
</div>
)}
<ChatInput
message={message}
onMessageChange={onMessageChange}
@ -223,26 +155,12 @@ const ChatPanel = ({
<span className="session-config-label">Agent</span>
<span className="session-config-value">{agentLabel}</span>
</div>
{currentAgentVersion && (
<div className="session-config-field">
<span className="session-config-label">Model</span>
<span className="session-config-value">{sessionModel || "-"}</span>
</div>
<div className="session-config-field">
<span className="session-config-label">Variant</span>
<span className="session-config-value">{sessionVariant || "-"}</span>
</div>
<div className="session-config-field">
<span className="session-config-label">Permission</span>
<span className="session-config-value">{sessionPermissionMode || "-"}</span>
</div>
<div className="session-config-field">
<span className="session-config-label">MCP Servers</span>
<span className="session-config-value">{sessionMcpServerCount}</span>
</div>
<div className="session-config-field">
<span className="session-config-label">Skills</span>
<span className="session-config-value">{sessionSkillSourceCount}</span>
<span className="session-config-label">Version</span>
<span className="session-config-value">{currentAgentVersion}</span>
</div>
)}
</div>
)}
</div>

13
frontend/packages/inspector/src/components/chat/messageUtils.ts
View file

@ -1,11 +1,10 @@
import type { UniversalItem } from "../../types/legacyApi";
import type { TimelineEntry } from "./types";
export const getMessageClass = (item: UniversalItem) => {
if (item.kind === "tool_call" || item.kind === "tool_result") return "tool";
if (item.kind === "system" || item.kind === "status") return "system";
if (item.role === "user") return "user";
if (item.role === "tool") return "tool";
if (item.role === "system") return "system";
export const getMessageClass = (entry: TimelineEntry) => {
if (entry.kind === "tool") return "tool";
if (entry.kind === "meta") return entry.meta?.severity === "error" ? "error" : "system";
if (entry.kind === "reasoning") return "assistant";
if (entry.role === "user") return "user";
return "assistant";
};

93
frontend/packages/inspector/src/components/chat/renderContentPart.tsx
View file

@ -1,93 +0,0 @@
import type { ContentPart } from "../../types/legacyApi";
import { formatJson } from "../../utils/format";
const renderContentPart = (part: ContentPart, index: number) => {
const partType = (part as { type?: string }).type ?? "unknown";
const key = `${partType}-${index}`;
switch (partType) {
case "text":
return (
<div key={key} className="part">
<div className="part-body">{(part as { text: string }).text}</div>
</div>
);
case "json":
return (
<div key={key} className="part">
<div className="part-title">json</div>
<pre className="code-block">{formatJson((part as { json: unknown }).json)}</pre>
</div>
);
case "tool_call": {
const { name, arguments: args, call_id } = part as {
name: string;
arguments: string;
call_id: string;
};
return (
<div key={key} className="part">
<div className="part-title">
tool call - {name}
{call_id ? ` - ${call_id}` : ""}
</div>
{args ? <pre className="code-block">{args}</pre> : <div className="muted">No arguments</div>}
</div>
);
}
case "tool_result": {
const { call_id, output } = part as { call_id: string; output: string };
return (
<div key={key} className="part">
<div className="part-title">tool result - {call_id}</div>
{output ? <pre className="code-block">{output}</pre> : <div className="muted">No output</div>}
</div>
);
}
case "file_ref": {
const { path, action, diff } = part as { path: string; action: string; diff?: string | null };
return (
<div key={key} className="part">
<div className="part-title">file - {action}</div>
<div className="part-body mono">{path}</div>
{diff && <pre className="code-block">{diff}</pre>}
</div>
);
}
case "reasoning": {
const { text, visibility } = part as { text: string; visibility: string };
return (
<div key={key} className="part">
<div className="part-title">reasoning - {visibility}</div>
<div className="part-body muted">{text}</div>
</div>
);
}
case "image": {
const { path, mime } = part as { path: string; mime?: string | null };
return (
<div key={key} className="part">
<div className="part-title">image {mime ? `- ${mime}` : ""}</div>
<div className="part-body mono">{path}</div>
</div>
);
}
case "status": {
const { label, detail } = part as { label: string; detail?: string | null };
return (
<div key={key} className="part">
<div className="part-title">status - {label}</div>
{detail && <div className="part-body">{detail}</div>}
</div>
);
}
default:
return (
<div key={key} className="part">
<div className="part-title">unknown</div>
<pre className="code-block">{formatJson(part)}</pre>
</div>
);
}
};
export default renderContentPart;

23
frontend/packages/inspector/src/components/chat/types.ts
View file

@ -1,14 +1,17 @@
import type { UniversalItem } from "../../types/legacyApi";
export type TimelineEntry = {
id: string;
kind: "item" | "meta";
kind: "message" | "tool" | "meta" | "reasoning";
time: string;
item?: UniversalItem;
deltaText?: string;
meta?: {
title: string;
detail?: string;
severity?: "info" | "error";
};
// For messages:
role?: "user" | "assistant";
text?: string;
// For tool calls:
toolName?: string;
toolInput?: string;
toolOutput?: string;
toolStatus?: string;
// For reasoning:
reasoning?: { text: string; visibility?: string };
// For meta:
meta?: { title: string; detail?: string; severity?: "info" | "error" };
};

10
frontend/packages/inspector/src/components/debug/AgentsTab.tsx
View file

@ -1,6 +1,8 @@
import { Download, Loader2, RefreshCw } from "lucide-react";
import { useState } from "react";
import type { AgentInfo, AgentModeInfo } from "../../types/legacyApi";
import type { AgentInfo } from "sandbox-agent";
type AgentModeInfo = { id: string; name: string; description: string };
import FeatureCoverageBadges from "../agents/FeatureCoverageBadges";
import { emptyFeatureCoverage } from "../../types/agents";
@ -52,9 +54,9 @@ const AgentsTab = ({
id,
installed: false,
credentialsAvailable: false,
version: undefined,
path: undefined,
capabilities: emptyFeatureCoverage
version: undefined as string | undefined,
path: undefined as string | undefined,
capabilities: emptyFeatureCoverage as AgentInfo["capabilities"],
}))).map((agent) => {
const isInstalling = installingAgent === agent.id;
return (

105
frontend/packages/inspector/src/components/debug/ApprovalsTab.tsx
View file

@ -1,105 +0,0 @@
import { HelpCircle, Shield } from "lucide-react";
import type { PermissionEventData, QuestionEventData } from "../../types/legacyApi";
import { formatJson } from "../../utils/format";
const ApprovalsTab = ({
questionRequests,
permissionRequests,
questionSelections,
onSelectQuestionOption,
onAnswerQuestion,
onRejectQuestion,
onReplyPermission
}: {
questionRequests: QuestionEventData[];
permissionRequests: PermissionEventData[];
questionSelections: Record<string, string[][]>;
onSelectQuestionOption: (requestId: string, optionLabel: string) => void;
onAnswerQuestion: (request: QuestionEventData) => void;
onRejectQuestion: (requestId: string) => void;
onReplyPermission: (requestId: string, reply: "once" | "always" | "reject") => void;
}) => {
return (
<>
{questionRequests.length === 0 && permissionRequests.length === 0 ? (
<div className="card-meta">No pending approvals.</div>
) : (
<>
{questionRequests.map((request) => {
const selections = questionSelections[request.question_id] ?? [];
const selected = selections[0] ?? [];
const answered = selected.length > 0;
return (
<div key={request.question_id} className="card">
<div className="card-header">
<span className="card-title">
<HelpCircle className="button-icon" style={{ marginRight: 6 }} />
Question
</span>
<span className="pill accent">Pending</span>
</div>
<div style={{ marginTop: 12 }}>
<div style={{ fontSize: 12, marginBottom: 8 }}>{request.prompt}</div>
<div className="option-list">
{request.options.map((option) => {
const isSelected = selected.includes(option);
return (
<label key={option} className="option-item">
<input
type="radio"
checked={isSelected}
onChange={() => onSelectQuestionOption(request.question_id, option)}
/>
<span>{option}</span>
</label>
);
})}
</div>
</div>
<div className="card-actions">
<button className="button success small" disabled={!answered} onClick={() => onAnswerQuestion(request)}>
Reply
</button>
<button className="button danger small" onClick={() => onRejectQuestion(request.question_id)}>
Reject
</button>
</div>
</div>
);
})}
{permissionRequests.map((request) => (
<div key={request.permission_id} className="card">
<div className="card-header">
<span className="card-title">
<Shield className="button-icon" style={{ marginRight: 6 }} />
Permission
</span>
<span className="pill accent">Pending</span>
</div>
<div className="card-meta" style={{ marginTop: 8 }}>
{request.action}
</div>
{request.metadata !== null && request.metadata !== undefined && (
<pre className="code-block">{formatJson(request.metadata)}</pre>
)}
<div className="card-actions">
<button className="button success small" onClick={() => onReplyPermission(request.permission_id, "once")}>
Allow Once
</button>
<button className="button secondary small" onClick={() => onReplyPermission(request.permission_id, "always")}>
Always
</button>
<button className="button danger small" onClick={() => onReplyPermission(request.permission_id, "reject")}>
Reject
</button>
</div>
</div>
))}
</>
)}
</>
);
};
export default ApprovalsTab;

38
frontend/packages/inspector/src/components/debug/DebugPanel.tsx
View file

@ -1,19 +1,21 @@
import { Cloud, PlayCircle, Terminal } from "lucide-react";
import type { AgentInfo, AgentModeInfo, UniversalEvent } from "../../types/legacyApi";
import { Cloud, PlayCircle, Server, Terminal, Wrench } from "lucide-react";
import type { AgentInfo, SandboxAgent, SessionEvent } from "sandbox-agent";
type AgentModeInfo = { id: string; name: string; description: string };
import AgentsTab from "./AgentsTab";
import EventsTab from "./EventsTab";
import McpTab from "./McpTab";
import SkillsTab from "./SkillsTab";
import RequestLogTab from "./RequestLogTab";
import type { RequestLog } from "../../types/requestLog";
export type DebugTab = "log" | "events" | "agents";
export type DebugTab = "log" | "events" | "agents" | "mcp" | "skills";
const DebugPanel = ({
debugTab,
onDebugTabChange,
events,
offset,
onResetEvents,
eventsError,
requestLog,
copiedLogId,
onClearRequestLog,
@ -24,14 +26,13 @@ const DebugPanel = ({
onRefreshAgents,
onInstallAgent,
agentsLoading,
agentsError
agentsError,
getClient,
}: {
debugTab: DebugTab;
onDebugTabChange: (tab: DebugTab) => void;
events: UniversalEvent[];
offset: number;
events: SessionEvent[];
onResetEvents: () => void;
eventsError: string | null;
requestLog: RequestLog[];
copiedLogId: number | null;
onClearRequestLog: () => void;
@ -43,6 +44,7 @@ const DebugPanel = ({
onInstallAgent: (agentId: string, reinstall: boolean) => Promise<void>;
agentsLoading: boolean;
agentsError: string | null;
getClient: () => SandboxAgent;
}) => {
return (
<div className="debug-panel">
@ -60,6 +62,14 @@ const DebugPanel = ({
<Cloud className="button-icon" style={{ marginRight: 4, width: 12, height: 12 }} />
Agents
</button>
<button className={`debug-tab ${debugTab === "mcp" ? "active" : ""}`} onClick={() => onDebugTabChange("mcp")}>
<Server className="button-icon" style={{ marginRight: 4, width: 12, height: 12 }} />
MCP
</button>
<button className={`debug-tab ${debugTab === "skills" ? "active" : ""}`} onClick={() => onDebugTabChange("skills")}>
<Wrench className="button-icon" style={{ marginRight: 4, width: 12, height: 12 }} />
Skills
</button>
</div>
<div className="debug-content">
@ -75,9 +85,7 @@ const DebugPanel = ({
{debugTab === "events" && (
<EventsTab
events={events}
offset={offset}
onClear={onResetEvents}
error={eventsError}
/>
)}
@ -92,6 +100,14 @@ const DebugPanel = ({
error={agentsError}
/>
)}
{debugTab === "mcp" && (
<McpTab getClient={getClient} />
)}
{debugTab === "skills" && (
<SkillsTab getClient={getClient} />
)}
</div>
</div>
);

153
frontend/packages/inspector/src/components/debug/EventsTab.tsx
View file

@ -1,19 +1,119 @@
import { ChevronDown, ChevronRight } from "lucide-react";
import {
Ban,
Bot,
Brain,
ChevronDown,
ChevronRight,
Circle,
CircleX,
Command,
CornerDownLeft,
FilePen,
FileText,
FolderOpen,
Hourglass,
KeyRound,
ListChecks,
MessageSquare,
Plug,
Radio,
ScrollText,
Settings,
ShieldCheck,
SquarePlus,
SquareTerminal,
ToggleLeft,
Trash2,
Unplug,
Wrench,
type LucideIcon,
} from "lucide-react";
import { useEffect, useState } from "react";
import type { UniversalEvent } from "../../types/legacyApi";
import type { SessionEvent } from "sandbox-agent";
import { formatJson, formatTime } from "../../utils/format";
import { getEventCategory, getEventClass, getEventIcon, getEventKey, getEventType } from "./eventUtils";
type EventIconInfo = { Icon: LucideIcon; category: string };
function getEventIcon(method: string, payload: Record<string, unknown>): EventIconInfo {
if (method === "session/update") {
const params = payload.params as Record<string, unknown> | undefined;
const update = params?.update as Record<string, unknown> | undefined;
const updateType = update?.sessionUpdate as string | undefined;
switch (updateType) {
case "user_message_chunk":
return { Icon: MessageSquare, category: "prompt" };
case "agent_message_chunk":
return { Icon: Bot, category: "update" };
case "agent_thought_chunk":
return { Icon: Brain, category: "update" };
case "tool_call":
case "tool_call_update":
return { Icon: Wrench, category: "tool" };
case "plan":
return { Icon: ListChecks, category: "config" };
case "available_commands_update":
return { Icon: Command, category: "config" };
case "current_mode_update":
return { Icon: ToggleLeft, category: "config" };
case "config_option_update":
return { Icon: Settings, category: "config" };
default:
return { Icon: Radio, category: "update" };
}
}
switch (method) {
case "initialize":
return { Icon: Plug, category: "connection" };
case "authenticate":
return { Icon: KeyRound, category: "connection" };
case "session/new":
return { Icon: SquarePlus, category: "session" };
case "session/load":
return { Icon: FolderOpen, category: "session" };
case "session/prompt":
return { Icon: MessageSquare, category: "prompt" };
case "session/cancel":
return { Icon: Ban, category: "cancel" };
case "session/set_mode":
return { Icon: ToggleLeft, category: "config" };
case "session/set_config_option":
return { Icon: Settings, category: "config" };
case "session/request_permission":
return { Icon: ShieldCheck, category: "permission" };
case "fs/read_text_file":
return { Icon: FileText, category: "filesystem" };
case "fs/write_text_file":
return { Icon: FilePen, category: "filesystem" };
case "terminal/create":
return { Icon: SquareTerminal, category: "terminal" };
case "terminal/kill":
return { Icon: CircleX, category: "terminal" };
case "terminal/output":
return { Icon: ScrollText, category: "terminal" };
case "terminal/release":
return { Icon: Trash2, category: "terminal" };
case "terminal/wait_for_exit":
return { Icon: Hourglass, category: "terminal" };
case "_sandboxagent/session/detach":
return { Icon: Unplug, category: "session" };
case "(response)":
return { Icon: CornerDownLeft, category: "response" };
default:
if (method.startsWith("_sandboxagent/")) {
return { Icon: Radio, category: "connection" };
}
return { Icon: Circle, category: "response" };
}
}
const EventsTab = ({
events,
offset,
onClear,
error
}: {
events: UniversalEvent[];
offset: number;
events: SessionEvent[];
onClear: () => void;
error: string | null;
}) => {
const [collapsedEvents, setCollapsedEvents] = useState<Record<string, boolean>>({});
const [copied, setCopied] = useState(false);
@ -55,10 +155,15 @@ const EventsTab = ({
}
}, [events.length]);
const getMethod = (event: SessionEvent): string => {
const payload = event.payload as Record<string, unknown>;
return typeof payload.method === "string" ? payload.method : "(response)";
};
return (
<>
<div className="inline-row" style={{ marginBottom: 12, justifyContent: "space-between" }}>
<span className="card-meta">Offset: {offset}</span>
<span className="card-meta">{events.length} events</span>
<div className="inline-row">
<button
type="button"
@ -75,26 +180,26 @@ const EventsTab = ({
</div>
</div>
{error && <div className="banner error">{error}</div>}
{events.length === 0 ? (
<div className="card-meta">
No events yet. Start streaming to receive events.
No events yet. Create a session and send a message.
</div>
) : (
<div className="event-list">
{[...events].reverse().map((event) => {
const type = getEventType(event);
const category = getEventCategory(type);
const eventClass = `${category} ${getEventClass(type)}`;
const eventKey = getEventKey(event);
const eventKey = event.id;
const isCollapsed = collapsedEvents[eventKey] ?? true;
const toggleCollapsed = () =>
setCollapsedEvents((prev) => ({
...prev,
[eventKey]: !(prev[eventKey] ?? true)
}));
const Icon = getEventIcon(type);
const method = getMethod(event);
const payload = event.payload as Record<string, unknown>;
const { Icon, category } = getEventIcon(method, payload);
const time = formatTime(new Date(event.createdAt).toISOString());
const senderClass = event.sender === "client" ? "client" : "agent";
return (
<div key={eventKey} className={`event-item ${isCollapsed ? "collapsed" : "expanded"}`}>
<button
@ -103,24 +208,26 @@ const EventsTab = ({
onClick={toggleCollapsed}
title={isCollapsed ? "Expand payload" : "Collapse payload"}
>
<span className={`event-icon ${eventClass}`}>
<span className={`event-icon ${category}`}>
<Icon size={14} />
</span>
<div className="event-summary-main">
<div className="event-title-row">
<span className={`event-type ${eventClass}`}>{type}</span>
<span className="event-time">{formatTime(event.time)}</span>
<span className={`event-type ${category}`}>{method}</span>
<span className={`pill ${senderClass === "client" ? "accent" : "success"}`}>
{event.sender}
</span>
<span className="event-time">{time}</span>
</div>
<div className="event-id">
Event #{event.event_id || event.sequence} - seq {event.sequence} - {event.source}
{event.synthetic ? " (synthetic)" : ""}
{event.id}
</div>
</div>
<span className="event-chevron">
{isCollapsed ? <ChevronRight size={16} /> : <ChevronDown size={16} />}
</span>
</button>
{!isCollapsed && <pre className="code-block event-payload">{formatJson(event.data)}</pre>}
{!isCollapsed && <pre className="code-block event-payload">{formatJson(event.payload)}</pre>}
</div>
);
})}

210
frontend/packages/inspector/src/components/debug/McpTab.tsx Normal file
View file

@ -0,0 +1,210 @@
import { FolderOpen, Loader2, Plus, Trash2 } from "lucide-react";
import { useCallback, useEffect, useState } from "react";
import type { SandboxAgent } from "sandbox-agent";
import { formatJson } from "../../utils/format";
type McpEntry = {
name: string;
config: Record<string, unknown>;
};
const McpTab = ({
getClient,
}: {
getClient: () => SandboxAgent;
}) => {
const [directory, setDirectory] = useState("/");
const [entries, setEntries] = useState<McpEntry[]>([]);
const [loading, setLoading] = useState(false);
const [error, setError] = useState<string | null>(null);
// Add/edit form state
const [editing, setEditing] = useState(false);
const [editName, setEditName] = useState("");
const [editJson, setEditJson] = useState("");
const [editError, setEditError] = useState<string | null>(null);
const [saving, setSaving] = useState(false);
const loadAll = useCallback(async (dir: string) => {
setLoading(true);
setError(null);
try {
const configPath = `${dir === "/" ? "" : dir}/.sandbox-agent/config/mcp.json`;
const bytes = await getClient().readFsFile({ path: configPath });
const text = new TextDecoder().decode(bytes);
if (!text.trim()) {
setEntries([]);
return;
}
const map = JSON.parse(text) as Record<string, Record<string, unknown>>;
setEntries(
Object.entries(map).map(([name, config]) => ({ name, config })),
);
} catch {
// File doesn't exist yet or is empty — that's fine
setEntries([]);
} finally {
setLoading(false);
}
}, [getClient]);
useEffect(() => {
loadAll(directory);
}, [directory, loadAll]);
const startAdd = () => {
setEditing(true);
setEditName("");
setEditJson('{\n "type": "local",\n "command": "npx",\n "args": ["@modelcontextprotocol/server-everything"]\n}');
setEditError(null);
};
const cancelEdit = () => {
setEditing(false);
setEditName("");
setEditJson("");
setEditError(null);
};
const save = async () => {
const name = editName.trim();
if (!name) {
setEditError("Name is required");
return;
}
let parsed: Record<string, unknown>;
try {
parsed = JSON.parse(editJson.trim());
if (!parsed || typeof parsed !== "object" || Array.isArray(parsed)) {
setEditError("Must be a JSON object");
return;
}
} catch {
setEditError("Invalid JSON");
return;
}
setSaving(true);
setEditError(null);
try {
await getClient().setMcpConfig(
{ directory, mcpName: name },
parsed as Parameters<SandboxAgent["setMcpConfig"]>[1],
);
cancelEdit();
await loadAll(directory);
} catch (err) {
setEditError(err instanceof Error ? err.message : "Failed to save");
} finally {
setSaving(false);
}
};
const remove = async (name: string) => {
try {
await getClient().deleteMcpConfig({ directory, mcpName: name });
await loadAll(directory);
} catch (err) {
setError(err instanceof Error ? err.message : "Failed to delete");
}
};
return (
<>
<div className="inline-row" style={{ marginBottom: 12, justifyContent: "space-between" }}>
<span className="card-meta">MCP Server Configuration</span>
<div className="inline-row">
{!editing && (
<button className="button secondary small" onClick={startAdd}>
<Plus className="button-icon" style={{ width: 12, height: 12 }} />
Add
</button>
)}
</div>
</div>
<div className="inline-row" style={{ marginBottom: 12, gap: 6 }}>
<FolderOpen size={14} className="muted" style={{ flexShrink: 0 }} />
<input
className="setup-input mono"
value={directory}
onChange={(e) => setDirectory(e.target.value)}
placeholder="/"
style={{ flex: 1, fontSize: 11 }}
/>
</div>
{error && <div className="banner error">{error}</div>}
{loading && <div className="card-meta">Loading...</div>}
{editing && (
<div className="card" style={{ marginBottom: 12 }}>
<div className="card-header">
<span className="card-title">
{editName ? `Edit: ${editName}` : "Add MCP Server"}
</span>
</div>
<div style={{ marginTop: 8 }}>
<input
className="setup-input"
value={editName}
onChange={(e) => { setEditName(e.target.value); setEditError(null); }}
placeholder="server-name"
style={{ marginBottom: 8, width: "100%", boxSizing: "border-box" }}
/>
<textarea
className="setup-input mono"
value={editJson}
onChange={(e) => { setEditJson(e.target.value); setEditError(null); }}
rows={6}
style={{ width: "100%", boxSizing: "border-box", fontFamily: "monospace", fontSize: 11 }}
/>
{editError && <div className="banner error" style={{ marginTop: 4 }}>{editError}</div>}
</div>
<div className="card-actions">
<button className="button primary small" onClick={save} disabled={saving}>
{saving ? <Loader2 className="button-icon spinner-icon" /> : null}
Save
</button>
<button className="button ghost small" onClick={cancelEdit}>
Cancel
</button>
</div>
</div>
)}
{entries.length === 0 && !editing && !loading && (
<div className="card-meta">
No MCP servers configured in this directory.
</div>
)}
{entries.map((entry) => (
<div key={entry.name} className="card" style={{ marginBottom: 8 }}>
<div className="card-header">
<span className="card-title">{entry.name}</span>
<div className="card-header-pills">
<span className="pill accent">
{(entry.config as { type?: string }).type ?? "unknown"}
</span>
<button
className="button ghost small"
onClick={() => remove(entry.name)}
title="Remove"
style={{ padding: "2px 4px" }}
>
<Trash2 size={12} />
</button>
</div>
</div>
<pre className="code-block" style={{ marginTop: 4, fontSize: 10 }}>
{formatJson(entry.config)}
</pre>
</div>
))}
</>
);
};
export default McpTab;

10
frontend/packages/inspector/src/components/debug/RequestLogTab.tsx
View file

@ -44,21 +44,21 @@ const RequestLogTab = ({
type="button"
onClick={() => hasDetails && toggleExpanded(entry.id)}
title={hasDetails ? (isExpanded ? "Collapse" : "Expand") : undefined}
style={{ cursor: hasDetails ? "pointer" : "default" }}
style={{ cursor: hasDetails ? "pointer" : "default", gridTemplateColumns: "1fr auto auto auto" }}
>
<div className="event-summary-main" style={{ flex: 1 }}>
<div className="event-summary-main">
<div className="event-title-row">
<span className="log-method">{entry.method}</span>
<span className="log-url text-truncate" style={{ flex: 1 }}>{entry.url}</span>
<span className={`log-status ${entry.status && entry.status < 400 ? "ok" : "error"}`}>
{entry.status || "ERR"}
</span>
</div>
<div className="event-id">
{entry.time}
{entry.error && ` - ${entry.error}`}
</div>
</div>
<span className={`log-status ${entry.status && entry.status < 400 ? "ok" : "error"}`}>
{entry.status || "ERR"}
</span>
<span
className="copy-button"
onClick={(e) => {

263
frontend/packages/inspector/src/components/debug/SkillsTab.tsx Normal file
View file

@ -0,0 +1,263 @@
import { FolderOpen, Loader2, Plus, Trash2 } from "lucide-react";
import { useCallback, useEffect, useState } from "react";
import type { SandboxAgent } from "sandbox-agent";
import { formatJson } from "../../utils/format";
type SkillEntry = {
name: string;
config: { sources: Array<{ source: string; type: string; ref?: string | null; subpath?: string | null; skills?: string[] | null }> };
};
const SkillsTab = ({
getClient,
}: {
getClient: () => SandboxAgent;
}) => {
const [directory, setDirectory] = useState("/");
const [entries, setEntries] = useState<SkillEntry[]>([]);
const [loading, setLoading] = useState(false);
const [error, setError] = useState<string | null>(null);
// Add form state
const [editing, setEditing] = useState(false);
const [editName, setEditName] = useState("");
const [editSource, setEditSource] = useState("");
const [editType, setEditType] = useState("github");
const [editRef, setEditRef] = useState("");
const [editSubpath, setEditSubpath] = useState("");
const [editSkills, setEditSkills] = useState("");
const [editError, setEditError] = useState<string | null>(null);
const [saving, setSaving] = useState(false);
const loadAll = useCallback(async (dir: string) => {
setLoading(true);
setError(null);
try {
const configPath = `${dir === "/" ? "" : dir}/.sandbox-agent/config/skills.json`;
const bytes = await getClient().readFsFile({ path: configPath });
const text = new TextDecoder().decode(bytes);
if (!text.trim()) {
setEntries([]);
return;
}
const map = JSON.parse(text) as Record<string, SkillEntry["config"]>;
setEntries(
Object.entries(map).map(([name, config]) => ({ name, config })),
);
} catch {
// File doesn't exist yet or is empty — that's fine
setEntries([]);
} finally {
setLoading(false);
}
}, [getClient]);
useEffect(() => {
loadAll(directory);
}, [directory, loadAll]);
const startAdd = () => {
setEditing(true);
setEditName("");
setEditSource("rivet-dev/skills");
setEditType("github");
setEditRef("");
setEditSubpath("");
setEditSkills("sandbox-agent");
setEditError(null);
};
const cancelEdit = () => {
setEditing(false);
setEditName("");
setEditSource("");
setEditType("github");
setEditRef("");
setEditSubpath("");
setEditSkills("");
setEditError(null);
};
const save = async () => {
const name = editName.trim();
if (!name) {
setEditError("Name is required");
return;
}
const source = editSource.trim();
if (!source) {
setEditError("Source is required");
return;
}
const skillEntry: SkillEntry["config"]["sources"][0] = {
source,
type: editType,
};
if (editRef.trim()) skillEntry.ref = editRef.trim();
if (editSubpath.trim()) skillEntry.subpath = editSubpath.trim();
const skillsList = editSkills.trim()
? editSkills.split(",").map((s) => s.trim()).filter(Boolean)
: null;
if (skillsList && skillsList.length > 0) skillEntry.skills = skillsList;
const config = { sources: [skillEntry] };
setSaving(true);
setEditError(null);
try {
await getClient().setSkillsConfig(
{ directory, skillName: name },
config,
);
cancelEdit();
await loadAll(directory);
} catch (err) {
setEditError(err instanceof Error ? err.message : "Failed to save");
} finally {
setSaving(false);
}
};
const remove = async (name: string) => {
try {
await getClient().deleteSkillsConfig({ directory, skillName: name });
await loadAll(directory);
} catch (err) {
setError(err instanceof Error ? err.message : "Failed to delete");
}
};
return (
<>
<div className="inline-row" style={{ marginBottom: 12, justifyContent: "space-between" }}>
<span className="card-meta">Skills Configuration</span>
<div className="inline-row">
{!editing && (
<button className="button secondary small" onClick={startAdd}>
<Plus className="button-icon" style={{ width: 12, height: 12 }} />
Add
</button>
)}
</div>
</div>
<div className="inline-row" style={{ marginBottom: 12, gap: 6 }}>
<FolderOpen size={14} className="muted" style={{ flexShrink: 0 }} />
<input
className="setup-input mono"
value={directory}
onChange={(e) => setDirectory(e.target.value)}
placeholder="/"
style={{ flex: 1, fontSize: 11 }}
/>
</div>
{error && <div className="banner error">{error}</div>}
{loading && <div className="card-meta">Loading...</div>}
{editing && (
<div className="card" style={{ marginBottom: 12 }}>
<div className="card-header">
<span className="card-title">Add Skill Source</span>
</div>
<div style={{ marginTop: 8 }}>
<input
className="setup-input"
value={editName}
onChange={(e) => { setEditName(e.target.value); setEditError(null); }}
placeholder="skill-name"
style={{ marginBottom: 6, width: "100%", boxSizing: "border-box" }}
/>
<div className="inline-row" style={{ marginBottom: 6, gap: 4 }}>
<select
className="setup-select"
value={editType}
onChange={(e) => setEditType(e.target.value)}
style={{ width: 90 }}
>
<option value="github">github</option>
<option value="local">local</option>
<option value="git">git</option>
</select>
<input
className="setup-input mono"
value={editSource}
onChange={(e) => { setEditSource(e.target.value); setEditError(null); }}
placeholder={editType === "github" ? "owner/repo" : editType === "local" ? "/path/to/skill" : "https://..."}
style={{ flex: 1 }}
/>
</div>
<input
className="setup-input"
value={editSkills}
onChange={(e) => setEditSkills(e.target.value)}
placeholder="Skills filter (comma-separated, optional)"
style={{ marginBottom: 6, width: "100%", boxSizing: "border-box" }}
/>
{editType !== "local" && (
<div className="inline-row" style={{ gap: 4 }}>
<input
className="setup-input mono"
value={editRef}
onChange={(e) => setEditRef(e.target.value)}
placeholder="Branch/tag (optional)"
style={{ flex: 1 }}
/>
<input
className="setup-input mono"
value={editSubpath}
onChange={(e) => setEditSubpath(e.target.value)}
placeholder="Subpath (optional)"
style={{ flex: 1 }}
/>
</div>
)}
{editError && <div className="banner error" style={{ marginTop: 4 }}>{editError}</div>}
</div>
<div className="card-actions">
<button className="button primary small" onClick={save} disabled={saving}>
{saving ? <Loader2 className="button-icon spinner-icon" /> : null}
Save
</button>
<button className="button ghost small" onClick={cancelEdit}>
Cancel
</button>
</div>
</div>
)}
{entries.length === 0 && !editing && !loading && (
<div className="card-meta">
No skills configured in this directory.
</div>
)}
{entries.map((entry) => (
<div key={entry.name} className="card" style={{ marginBottom: 8 }}>
<div className="card-header">
<span className="card-title">{entry.name}</span>
<div className="card-header-pills">
<span className="pill accent">
{entry.config.sources.length} source{entry.config.sources.length !== 1 ? "s" : ""}
</span>
<button
className="button ghost small"
onClick={() => remove(entry.name)}
title="Remove"
style={{ padding: "2px 4px" }}
>
<Trash2 size={12} />
</button>
</div>
</div>
<pre className="code-block" style={{ marginTop: 4, fontSize: 10 }}>
{formatJson(entry.config)}
</pre>
</div>
))}
</>
);
};
export default SkillsTab;

110
frontend/packages/inspector/src/components/debug/eventUtils.ts
View file

@ -1,110 +0,0 @@
import {
Activity,
AlertTriangle,
Brain,
CheckCircle,
FileDiff,
HelpCircle,
Info,
MessageSquare,
PauseCircle,
PlayCircle,
Shield,
Terminal,
Wrench,
Zap
} from "lucide-react";
import type { UniversalEvent } from "../../types/legacyApi";
export const getEventType = (event: UniversalEvent) => event.type;
export const getEventKey = (event: UniversalEvent) =>
event.event_id ? `id:${event.event_id}` : `seq:${event.sequence}`;
export const getEventCategory = (type: string) => type.split(".")[0] ?? type;
export const getEventClass = (type: string) => type.replace(/\./g, "-");
export const getEventIcon = (type: string) => {
switch (type) {
// ACP session update events
case "acp.agent_message_chunk":
return MessageSquare;
case "acp.user_message_chunk":
return MessageSquare;
case "acp.agent_thought_chunk":
return Brain;
case "acp.tool_call":
return Wrench;
case "acp.tool_call_update":
return Activity;
case "acp.plan":
return FileDiff;
case "acp.session_info_update":
return Info;
case "acp.usage_update":
return Info;
case "acp.current_mode_update":
return Info;
case "acp.config_option_update":
return Info;
case "acp.available_commands_update":
return Terminal;
// Inspector lifecycle events
case "inspector.turn_started":
return PlayCircle;
case "inspector.turn_ended":
return PauseCircle;
case "inspector.user_message":
return MessageSquare;
// Session lifecycle (inspector-emitted)
case "session.started":
return PlayCircle;
case "session.ended":
return PauseCircle;
// Legacy synthetic events
case "turn.started":
return PlayCircle;
case "turn.ended":
return PauseCircle;
case "item.started":
return MessageSquare;
case "item.delta":
return Activity;
case "item.completed":
return CheckCircle;
// Approval events
case "question.requested":
return HelpCircle;
case "question.resolved":
return CheckCircle;
case "permission.requested":
return Shield;
case "permission.resolved":
return CheckCircle;
// Error events
case "error":
return AlertTriangle;
case "agent.unparsed":
return Brain;
default:
if (type.startsWith("acp.")) return Zap;
if (type.startsWith("inspector.")) return Info;
if (type.startsWith("item.")) return MessageSquare;
if (type.startsWith("session.")) return PlayCircle;
if (type.startsWith("error")) return AlertTriangle;
if (type.startsWith("agent.")) return Brain;
if (type.startsWith("question.")) return HelpCircle;
if (type.startsWith("permission.")) return Shield;
if (type.startsWith("file.")) return FileDiff;
if (type.startsWith("command.")) return Terminal;
if (type.startsWith("tool.")) return Wrench;
return Zap;
}
};

790
frontend/packages/inspector/src/lib/legacyClient.ts
View file

@ -1,790 +0,0 @@
import {
SandboxAgent,
type PermissionOption,
type RequestPermissionRequest,
type RequestPermissionResponse,
type SandboxAgentAcpClient,
type SandboxAgentConnectOptions,
type SessionNotification,
} from "sandbox-agent";
import type {
AgentInfo,
AgentModelInfo,
AgentModeInfo,
AgentModelsResponse,
AgentModesResponse,
CreateSessionRequest,
EventsQuery,
EventsResponse,
MessageRequest,
PermissionEventData,
PermissionReplyRequest,
QuestionEventData,
QuestionReplyRequest,
SessionInfo,
SessionListResponse,
TurnStreamQuery,
UniversalEvent,
} from "../types/legacyApi";
type PendingPermission = {
request: RequestPermissionRequest;
resolve: (response: RequestPermissionResponse) => void;
autoEndTurnOnResolve?: boolean;
};
type PendingQuestion = {
prompt: string;
options: string[];
autoEndTurnOnResolve?: boolean;
};
type RuntimeSession = {
aliasSessionId: string;
realSessionId: string;
agent: string;
connection: SandboxAgentAcpClient;
events: UniversalEvent[];
nextSequence: number;
listeners: Set<(event: UniversalEvent) => void>;
info: SessionInfo;
pendingPermissions: Map<string, PendingPermission>;
pendingQuestions: Map<string, PendingQuestion>;
};
const TDOO_PERMISSION_MODE =
"TDOO: ACP permission mode preconfiguration is not implemented in inspector compatibility.";
const TDOO_VARIANT =
"TDOO: ACP session variants are not implemented in inspector compatibility.";
const TDOO_SKILLS =
"TDOO: ACP skills source configuration is not implemented in inspector compatibility.";
const TDOO_MODE_DISCOVERY =
"TDOO: ACP mode discovery before session creation is not implemented; returning cached/empty modes.";
const TDOO_MODEL_DISCOVERY =
"TDOO: ACP model discovery before session creation is not implemented; returning cached/empty models.";
export class InspectorLegacyClient {
private readonly base: SandboxAgent;
private readonly sessions = new Map<string, RuntimeSession>();
private readonly aliasByRealSessionId = new Map<string, string>();
private readonly modeCache = new Map<string, AgentModeInfo[]>();
private readonly modelCache = new Map<string, AgentModelsResponse>();
private permissionCounter = 0;
private constructor(base: SandboxAgent) {
this.base = base;
}
static async connect(options: SandboxAgentConnectOptions): Promise<InspectorLegacyClient> {
const base = await SandboxAgent.connect(options);
return new InspectorLegacyClient(base);
}
async getHealth() {
return this.base.getHealth();
}
async listAgents(): Promise<{ agents: AgentInfo[] }> {
const response = await this.base.listAgents();
return {
agents: response.agents.map((agent) => {
const installed =
agent.agent_process_installed &&
(!agent.native_required || agent.native_installed);
return {
id: agent.id,
installed,
credentialsAvailable: true,
version: agent.agent_process_version ?? agent.native_version ?? null,
path: null,
capabilities: {
unstable_methods: agent.capabilities.unstable_methods,
},
native_required: agent.native_required,
native_installed: agent.native_installed,
native_version: agent.native_version,
agent_process_installed: agent.agent_process_installed,
agent_process_source: agent.agent_process_source,
agent_process_version: agent.agent_process_version,
};
}),
};
}
async installAgent(agent: string, request: { reinstall?: boolean } = {}) {
return this.base.installAgent(agent, request);
}
async getAgentModes(agentId: string): Promise<AgentModesResponse> {
const modes = this.modeCache.get(agentId);
if (modes) {
return { modes };
}
console.warn(TDOO_MODE_DISCOVERY);
return { modes: [] };
}
async getAgentModels(agentId: string): Promise<AgentModelsResponse> {
const models = this.modelCache.get(agentId);
if (models) {
return models;
}
console.warn(TDOO_MODEL_DISCOVERY);
return { models: [], defaultModel: null };
}
async createSession(aliasSessionId: string, request: CreateSessionRequest): Promise<void> {
await this.terminateSession(aliasSessionId).catch(() => {
// Ignore if it doesn't exist yet.
});
const acp = await this.base.createAcpClient({
agent: request.agent,
client: {
sessionUpdate: async (notification) => {
this.handleSessionUpdate(notification);
},
requestPermission: async (permissionRequest) => {
return this.handlePermissionRequest(permissionRequest);
},
},
});
await acp.initialize();
const created = await acp.newSession({
cwd: "/",
mcpServers: convertMcpConfig(request.mcp ?? {}),
});
if (created.modes?.availableModes) {
this.modeCache.set(
request.agent,
created.modes.availableModes.map((mode) => ({
id: mode.id,
name: mode.name,
description: mode.description ?? undefined,
})),
);
}
if (created.models?.availableModels) {
this.modelCache.set(request.agent, {
models: created.models.availableModels.map((model) => ({
id: model.modelId,
name: model.name,
description: model.description ?? undefined,
})),
defaultModel: created.models.currentModelId ?? null,
});
}
const runtime: RuntimeSession = {
aliasSessionId,
realSessionId: created.sessionId,
agent: request.agent,
connection: acp,
events: [],
nextSequence: 1,
listeners: new Set(),
info: {
sessionId: aliasSessionId,
agent: request.agent,
eventCount: 0,
ended: false,
model: request.model ?? null,
variant: request.variant ?? null,
permissionMode: request.permissionMode ?? null,
mcp: request.mcp,
skills: request.skills,
},
pendingPermissions: new Map(),
pendingQuestions: new Map(),
};
this.sessions.set(aliasSessionId, runtime);
this.aliasByRealSessionId.set(created.sessionId, aliasSessionId);
if (request.agentMode) {
try {
await acp.setSessionMode({ sessionId: created.sessionId, modeId: request.agentMode });
} catch {
this.emitError(aliasSessionId, `TDOO: Unable to apply mode \"${request.agentMode}\" via ACP.`);
}
}
if (request.model) {
try {
await acp.unstableSetSessionModel({
sessionId: created.sessionId,
modelId: request.model,
});
} catch {
this.emitError(aliasSessionId, `TDOO: Unable to apply model \"${request.model}\" via ACP.`);
}
}
if (request.permissionMode) {
this.emitError(aliasSessionId, TDOO_PERMISSION_MODE);
}
if (request.variant) {
this.emitError(aliasSessionId, TDOO_VARIANT);
}
if (request.skills?.sources && request.skills.sources.length > 0) {
this.emitError(aliasSessionId, TDOO_SKILLS);
}
this.emitEvent(aliasSessionId, "session.started", {
session_id: aliasSessionId,
agent: request.agent,
});
}
async listSessions(): Promise<SessionListResponse> {
const sessions = Array.from(this.sessions.values()).map((session) => {
return {
...session.info,
eventCount: session.events.length,
};
});
return { sessions };
}
async postMessage(sessionId: string, request: MessageRequest): Promise<void> {
const runtime = this.requireActiveSession(sessionId);
const message = request.message.trim();
if (!message) {
return;
}
this.emitEvent(sessionId, "inspector.turn_started", {
session_id: sessionId,
});
this.emitEvent(sessionId, "inspector.user_message", {
session_id: sessionId,
text: message,
});
try {
await runtime.connection.prompt({
sessionId: runtime.realSessionId,
prompt: [{ type: "text", text: message }],
});
} catch (error) {
const detail = error instanceof Error ? error.message : "prompt failed";
this.emitError(sessionId, detail);
throw error;
} finally {
this.emitEvent(sessionId, "inspector.turn_ended", {
session_id: sessionId,
});
}
}
async getEvents(sessionId: string, query: EventsQuery = {}): Promise<EventsResponse> {
const runtime = this.requireSession(sessionId);
const offset = query.offset ?? 0;
const limit = query.limit ?? 200;
const events = runtime.events.filter((event) => event.sequence > offset).slice(0, limit);
return { events };
}
async *streamEvents(
sessionId: string,
query: EventsQuery = {},
signal?: AbortSignal,
): AsyncIterable<UniversalEvent> {
const runtime = this.requireSession(sessionId);
let cursor = query.offset ?? 0;
for (const event of runtime.events) {
if (event.sequence <= cursor) {
continue;
}
cursor = event.sequence;
yield event;
}
const queue: UniversalEvent[] = [];
let wake: (() => void) | null = null;
const listener = (event: UniversalEvent) => {
if (event.sequence <= cursor) {
return;
}
queue.push(event);
if (wake) {
wake();
wake = null;
}
};
runtime.listeners.add(listener);
try {
while (!signal?.aborted) {
if (queue.length === 0) {
await waitForSignalOrEvent(signal, () => {
wake = () => {};
return new Promise<void>((resolve) => {
wake = resolve;
});
});
continue;
}
const next = queue.shift();
if (!next) {
continue;
}
cursor = next.sequence;
yield next;
}
} finally {
runtime.listeners.delete(listener);
}
}
async *streamTurn(
sessionId: string,
request: MessageRequest,
_query?: TurnStreamQuery,
signal?: AbortSignal,
): AsyncIterable<UniversalEvent> {
if (signal?.aborted) {
return;
}
const runtime = this.requireActiveSession(sessionId);
let cursor = runtime.nextSequence - 1;
const queue: UniversalEvent[] = [];
let wake: (() => void) | null = null;
let promptDone = false;
let promptError: unknown = null;
const notify = () => {
if (wake) {
wake();
wake = null;
}
};
const listener = (event: UniversalEvent) => {
if (event.sequence <= cursor) {
return;
}
queue.push(event);
notify();
};
runtime.listeners.add(listener);
const promptPromise = this.postMessage(sessionId, request)
.catch((error) => {
promptError = error;
})
.finally(() => {
promptDone = true;
notify();
});
try {
while (!signal?.aborted) {
if (queue.length === 0) {
if (promptDone) {
break;
}
await waitForSignalOrEvent(signal, () => {
wake = () => {};
return new Promise<void>((resolve) => {
wake = resolve;
});
});
continue;
}
const next = queue.shift();
if (!next) {
continue;
}
cursor = next.sequence;
yield next;
}
} finally {
runtime.listeners.delete(listener);
}
await promptPromise;
if (promptError) {
throw promptError;
}
}
async replyQuestion(
sessionId: string,
questionId: string,
request: QuestionReplyRequest,
): Promise<void> {
const runtime = this.requireSession(sessionId);
const pending = runtime.pendingQuestions.get(questionId);
if (!pending) {
throw new Error("TDOO: Question request no longer pending.");
}
runtime.pendingQuestions.delete(questionId);
const response = request.answers?.[0]?.[0] ?? null;
const resolved: QuestionEventData & { response?: string | null } = {
question_id: questionId,
status: "resolved",
prompt: pending.prompt,
options: pending.options,
response,
};
this.emitEvent(sessionId, "question.resolved", resolved);
if (pending.autoEndTurnOnResolve) {
this.emitEvent(sessionId, "turn.ended", { session_id: sessionId });
}
}
async rejectQuestion(sessionId: string, questionId: string): Promise<void> {
const runtime = this.requireSession(sessionId);
const pending = runtime.pendingQuestions.get(questionId);
if (!pending) {
throw new Error("TDOO: Question request no longer pending.");
}
runtime.pendingQuestions.delete(questionId);
const resolved: QuestionEventData & { response?: string | null } = {
question_id: questionId,
status: "resolved",
prompt: pending.prompt,
options: pending.options,
response: null,
};
this.emitEvent(sessionId, "question.resolved", resolved);
if (pending.autoEndTurnOnResolve) {
this.emitEvent(sessionId, "turn.ended", { session_id: sessionId });
}
}
async replyPermission(
sessionId: string,
permissionId: string,
request: PermissionReplyRequest,
): Promise<void> {
const runtime = this.requireSession(sessionId);
const pending = runtime.pendingPermissions.get(permissionId);
if (!pending) {
throw new Error("TDOO: Permission request no longer pending.");
}
const optionId = selectPermissionOption(pending.request.options, request.reply);
const response: RequestPermissionResponse = optionId
? {
outcome: {
outcome: "selected",
optionId,
},
}
: {
outcome: {
outcome: "cancelled",
},
};
pending.resolve(response);
runtime.pendingPermissions.delete(permissionId);
const action = pending.request.toolCall.title ?? pending.request.toolCall.kind ?? "permission";
const resolved: PermissionEventData = {
permission_id: permissionId,
status: "resolved",
action,
metadata: {
reply: request.reply,
},
};
this.emitEvent(sessionId, "permission.resolved", resolved);
if (pending.autoEndTurnOnResolve) {
this.emitEvent(sessionId, "turn.ended", { session_id: sessionId });
}
}
async terminateSession(sessionId: string): Promise<void> {
const runtime = this.sessions.get(sessionId);
if (!runtime) {
return;
}
this.emitEvent(sessionId, "session.ended", {
reason: "terminated_by_user",
terminated_by: "inspector",
});
runtime.info.ended = true;
for (const pending of runtime.pendingPermissions.values()) {
pending.resolve({
outcome: {
outcome: "cancelled",
},
});
}
runtime.pendingPermissions.clear();
runtime.pendingQuestions.clear();
try {
await runtime.connection.close();
} catch {
// Best-effort close.
}
this.aliasByRealSessionId.delete(runtime.realSessionId);
}
async dispose(): Promise<void> {
for (const sessionId of Array.from(this.sessions.keys())) {
await this.terminateSession(sessionId);
}
await this.base.dispose();
}
private handleSessionUpdate(notification: SessionNotification): void {
const aliasSessionId = this.aliasByRealSessionId.get(notification.sessionId);
if (!aliasSessionId) {
return;
}
const runtime = this.sessions.get(aliasSessionId);
if (!runtime || runtime.info.ended) {
return;
}
const update = notification.update;
// Still handle session_info_update for sidebar metadata
if (update.sessionUpdate === "session_info_update") {
runtime.info.title = update.title ?? runtime.info.title;
runtime.info.updatedAt = update.updatedAt ?? runtime.info.updatedAt;
}
// Emit the raw notification as the event data, using the ACP discriminator as the type
this.emitEvent(aliasSessionId, `acp.${update.sessionUpdate}`, notification);
}
private async handlePermissionRequest(
request: RequestPermissionRequest,
): Promise<RequestPermissionResponse> {
const aliasSessionId = this.aliasByRealSessionId.get(request.sessionId);
if (!aliasSessionId) {
return {
outcome: {
outcome: "cancelled",
},
};
}
const runtime = this.sessions.get(aliasSessionId);
if (!runtime || runtime.info.ended) {
return {
outcome: {
outcome: "cancelled",
},
};
}
this.permissionCounter += 1;
const permissionId = `permission-${this.permissionCounter}`;
const action = request.toolCall.title ?? request.toolCall.kind ?? "permission";
const pendingEvent: PermissionEventData = {
permission_id: permissionId,
status: "requested",
action,
metadata: request,
};
this.emitEvent(aliasSessionId, "permission.requested", pendingEvent);
return await new Promise<RequestPermissionResponse>((resolve) => {
runtime.pendingPermissions.set(permissionId, { request, resolve });
});
}
private emitError(sessionId: string, message: string): void {
this.emitEvent(sessionId, "error", {
message,
});
}
private emitEvent(sessionId: string, type: string, data: unknown): void {
const runtime = this.sessions.get(sessionId);
if (!runtime) {
return;
}
const event: UniversalEvent = {
event_id: `${sessionId}-${runtime.nextSequence}`,
sequence: runtime.nextSequence,
type,
source: "inspector.acp",
time: new Date().toISOString(),
synthetic: true,
data,
};
runtime.nextSequence += 1;
runtime.events.push(event);
runtime.info.eventCount = runtime.events.length;
for (const listener of runtime.listeners) {
listener(event);
}
}
private requireSession(sessionId: string): RuntimeSession {
const runtime = this.sessions.get(sessionId);
if (!runtime) {
throw new Error(`Session not found: ${sessionId}`);
}
return runtime;
}
private requireActiveSession(sessionId: string): RuntimeSession {
const runtime = this.requireSession(sessionId);
if (runtime.info.ended) {
throw new Error(`Session ended: ${sessionId}`);
}
return runtime;
}
}
const convertMcpConfig = (mcp: Record<string, unknown>) => {
return Object.entries(mcp)
.map(([name, config]) => {
if (!config || typeof config !== "object") {
return null;
}
const value = config as Record<string, unknown>;
const type = value.type;
if (type === "local") {
const commandValue = value.command;
const argsValue = value.args;
let command = "";
let args: string[] = [];
if (Array.isArray(commandValue) && commandValue.length > 0) {
command = String(commandValue[0] ?? "");
args = commandValue.slice(1).map((part) => String(part));
} else if (typeof commandValue === "string") {
command = commandValue;
}
if (Array.isArray(argsValue)) {
args = argsValue.map((part) => String(part));
}
const envObject =
value.env && typeof value.env === "object" ? (value.env as Record<string, unknown>) : {};
const env = Object.entries(envObject).map(([envName, envValue]) => ({
name: envName,
value: String(envValue),
}));
return {
name,
command,
args,
env,
};
}
if (type === "remote") {
const headersObject =
value.headers && typeof value.headers === "object"
? (value.headers as Record<string, unknown>)
: {};
const headers = Object.entries(headersObject).map(([headerName, headerValue]) => ({
name: headerName,
value: String(headerValue),
}));
return {
type: "http" as const,
name,
url: String(value.url ?? ""),
headers,
};
}
return null;
})
.filter((entry): entry is NonNullable<typeof entry> => entry !== null);
};
const selectPermissionOption = (
options: PermissionOption[],
reply: PermissionReplyRequest["reply"],
): string | null => {
const pick = (...kinds: PermissionOption["kind"][]) => {
return options.find((option) => kinds.includes(option.kind))?.optionId ?? null;
};
if (reply === "always") {
return pick("allow_always", "allow_once");
}
if (reply === "once") {
return pick("allow_once", "allow_always");
}
return pick("reject_once", "reject_always");
};
const waitForSignalOrEvent = async (
signal: AbortSignal | undefined,
createWaitPromise: () => Promise<void>,
) => {
if (signal?.aborted) {
return;
}
await new Promise<void>((resolve) => {
let done = false;
const finish = () => {
if (done) {
return;
}
done = true;
if (signal) {
signal.removeEventListener("abort", onAbort);
}
resolve();
};
const onAbort = () => finish();
if (signal) {
signal.addEventListener("abort", onAbort, { once: true });
}
createWaitPromise().then(finish).catch(finish);
});
};

145
frontend/packages/inspector/src/types/legacyApi.ts
View file

@ -1,145 +0,0 @@
export type SkillSourceType = "github" | "local" | "git";
export type SkillSource = {
type: SkillSourceType;
source: string;
skills?: string[];
ref?: string;
subpath?: string;
};
export type CreateSessionRequest = {
agent: string;
agentMode?: string;
permissionMode?: string;
model?: string;
variant?: string;
mcp?: Record<string, unknown>;
skills?: {
sources: SkillSource[];
};
};
export type AgentModeInfo = {
id: string;
name?: string;
description?: string;
};
export type AgentModelInfo = {
id: string;
name?: string;
description?: string;
variants?: string[];
};
export type AgentInfo = {
id: string;
installed: boolean;
credentialsAvailable: boolean;
version?: string | null;
path?: string | null;
capabilities: Record<string, boolean | undefined>;
native_required?: boolean;
native_installed?: boolean;
native_version?: string | null;
agent_process_installed?: boolean;
agent_process_source?: string | null;
agent_process_version?: string | null;
};
export type ContentPart = {
type?: string;
[key: string]: unknown;
};
export type UniversalItem = {
item_id: string;
native_item_id?: string | null;
parent_id?: string | null;
kind: string;
role?: string | null;
content?: ContentPart[];
status?: string | null;
[key: string]: unknown;
};
export type UniversalEvent = {
event_id: string;
sequence: number;
type: string;
source: string;
time: string;
synthetic?: boolean;
data: unknown;
[key: string]: unknown;
};
export type PermissionEventData = {
permission_id: string;
status: "requested" | "resolved";
action: string;
metadata?: unknown;
};
export type QuestionEventData = {
question_id: string;
status: "requested" | "resolved";
prompt: string;
options: string[];
};
export type SessionInfo = {
sessionId: string;
agent: string;
eventCount: number;
ended?: boolean;
model?: string | null;
variant?: string | null;
permissionMode?: string | null;
mcp?: Record<string, unknown>;
skills?: {
sources?: SkillSource[];
};
title?: string | null;
updatedAt?: string | null;
};
export type EventsQuery = {
offset?: number;
limit?: number;
includeRaw?: boolean;
};
export type EventsResponse = {
events: UniversalEvent[];
};
export type SessionListResponse = {
sessions: SessionInfo[];
};
export type AgentModesResponse = {
modes: AgentModeInfo[];
};
export type AgentModelsResponse = {
models: AgentModelInfo[];
defaultModel?: string | null;
};
export type MessageRequest = {
message: string;
};
export type TurnStreamQuery = {
includeRaw?: boolean;
};
export type PermissionReplyRequest = {
reply: "once" | "always" | "reject";
};
export type QuestionReplyRequest = {
answers: string[][];
};

2
frontend/packages/inspector/vite.config.ts
View file

@ -7,7 +7,7 @@ export default defineConfig(({ command }) => ({
server: {
port: 5173,
proxy: {
"/v2": {
"/v1": {
target: "http://localhost:2468",
changeOrigin: true,
},

2
frontend/packages/website/src/components/FAQ.tsx
View file

@ -18,7 +18,7 @@ const faqs = [
{
question: 'How is session data persisted?',
answer:
"This SDK does not handle persisting session data. In v2, traffic is ACP JSON-RPC over <code>/v2/rpc</code>; persist envelopes in your own storage if you need replay or auditing.",
"This SDK does not handle persisting session data. Events stream in a universal JSON schema that you can persist anywhere. Consider using Postgres or <a href='https://rivet.gg' target='_blank' rel='noopener noreferrer' class='text-orange-400 hover:underline'>Rivet Actors</a> for data persistence.",
},
{
question: 'Can I run this locally or does it require a sandbox provider?',

4
frontend/packages/website/src/components/GetStarted.tsx
View file

@ -86,7 +86,7 @@ function SdkCodeHighlighted() {
);
}
const sandboxCommand = `curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/install.sh | sh`;
const sandboxCommand = `curl -fsSL https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh | sh`;
const sourceCommands = `git clone https://github.com/rivet-dev/sandbox-agent
cd sandbox-agent
@ -172,7 +172,7 @@ export function GetStarted() {
<span className="text-zinc-300">curl -fsSL \</span>
{"\n"}
<span className="text-zinc-300">{" "}</span>
<span className="text-green-400">https://releases.rivet.dev/sandbox-agent/latest/install.sh</span>
<span className="text-green-400">https://releases.rivet.dev/sandbox-agent/0.2.x/install.sh</span>
<span className="text-zinc-300"> | </span>
<span className="text-blue-400">sh</span>
</code>

6
research/acp/00-delete-first.md
View file

@ -1,6 +1,6 @@
# Delete Or Comment Out First
This is the initial, deliberate teardown list before building ACP-native v2.
This is the initial, deliberate teardown list before building ACP-native v1.
## Hard delete first (in-house protocol types and converters)
@ -39,7 +39,7 @@ This is the initial, deliberate teardown list before building ACP-native v2.
Rationale: this layer is based on current v1 session/event model. Comment it out/disable it during ACP core implementation to avoid coupling and drift.
Important: OpenCode <-> ACP support is still required, but it is explicitly reintroduced in Phase 7 after ACP v2 core transport/runtime are stable.
Important: OpenCode <-> ACP support is still required, but it is explicitly reintroduced in Phase 7 after ACP v1 core transport/runtime are stable.
## Tests to remove or disable with v1
@ -50,4 +50,4 @@ Important: OpenCode <-> ACP support is still required, but it is explicitly rein
- `server/packages/sandbox-agent/tests/sessions.rs`
- `server/packages/sandbox-agent/tests/agent_flows.rs`
Replace with ACP-native contract tests in v2.
Replace with ACP-native contract tests in v1.

8
research/acp/README.md
View file

@ -1,13 +1,13 @@
# ACP Migration Research
This folder captures the v2 migration plan from the current in-house protocol to ACP-first architecture.
This folder captures the v1 migration plan from the current in-house protocol to ACP-first architecture.
## Files
- `research/acp/00-delete-first.md`: delete/comment-out-first inventory for the rewrite kickoff.
- `research/acp/acp-notes.md`: ACP protocol notes extracted from `~/misc/acp-docs`.
- `research/acp/acp-over-http-findings.md`: field research from ACP Zulip thread on real ACP-over-HTTP transport patterns and recommendations.
- `research/acp/spec.md`: proposed v2 protocol/transport spec (ACP over HTTP).
- `research/acp/spec.md`: proposed v1 protocol/transport spec (ACP over HTTP).
- `research/acp/v1-schema-to-acp-mapping.md`: exhaustive 1:1 mapping of all current v1 endpoints/events into ACP methods, notifications, responses, and `_meta` extensions.
- `research/acp/rfds-vs-extensions.md`: simple list of which gaps should be raised as ACP RFDs vs remain product-specific extensions.
- `research/acp/migration-steps.md`: concrete implementation phases and execution checklist.
@ -35,7 +35,7 @@ This folder captures the v2 migration plan from the current in-house protocol to
## Important context
- ACP stable transport is stdio; streamable HTTP is still draft in ACP docs.
- v2 in this repo is intentionally breaking and ACP-native.
- v1 is removed in v2 and returns HTTP 410 on `/v1/*`.
- v1 in this repo is intentionally breaking and ACP-native.
- v1 is removed in v1 and returns HTTP 410 on `/v1/*`.
- `/opencode/*` is disabled during ACP core phases and re-enabled in the dedicated bridge phase.
- Keep `research/acp/friction.md` current as issues/ambiguities are discovered.

2
research/acp/acp-notes.md
View file

@ -63,4 +63,4 @@ From ACP docs agent list:
Gap to confirm for launch scope:
- Amp is not currently listed in ACP docs as a native ACP agent or published agent process.
- We need an explicit product decision: block Amp in v2 launch or provide/build an ACP agent process.
- We need an explicit product decision: block Amp in v1 launch or provide/build an ACP agent process.

16
research/acp/acp-over-http-findings.md
View file

@ -53,14 +53,14 @@ Why it matters:
- Lower complexity for bidirectional ACP semantics.
- But less aligned with strict HTTP-only environments without additional gatewaying.
## Recommended options for our v2
## Recommended options for our v1
## Option A (recommended): Streamable HTTP as canonical v2 transport
## Option A (recommended): Streamable HTTP as canonical v1 transport
Implement ACP over:
- `POST /v2/rpc`
- `GET /v2/rpc` (SSE, optional but recommended)
- `DELETE /v2/rpc`
- `POST /v1/rpc`
- `GET /v1/rpc` (SSE, optional but recommended)
- `DELETE /v1/rpc`
Profile:
- Keep JSON-RPC payloads pure ACP.
@ -86,14 +86,14 @@ Pros:
- Potentially simpler core runtime behavior.
Cons:
- Less direct fit to your immediate "ACP over HTTP v2 API" objective.
- Less direct fit to your immediate "ACP over HTTP v1 API" objective.
- Requires and maintains a translation layer from day one.
## Recommendation
Choose Option A for v2 launch and keep Option B as a later optimization path if operational pain appears.
Choose Option A for v1 launch and keep Option B as a later optimization path if operational pain appears.
Rationale:
- It matches current product direction.
- It aligns with concrete ecosystem work already visible (Goose Streamable HTTP).
- It can still preserve a future WebSocket backend if needed later, without changing v2 public semantics.
- It can still preserve a future WebSocket backend if needed later, without changing v1 public semantics.

4
research/acp/extensibility-status.md
View file

@ -13,7 +13,7 @@ Primary references:
## 1) Status Matrix (Session-Centric)
| v1 capability (session-related) | ACP stable | ACP unstable | Status in v2 | Recommendation |
| v1 capability (session-related) | ACP stable | ACP unstable | Status in v1 | Recommendation |
|---|---|---|---|---|
| Create session | `session/new` | N/A | Covered | Use ACP standard only. |
| Load/replay prior session | `session/load` (capability-gated) | N/A | Covered when agent process supports `loadSession` | Keep standard behavior. |
@ -111,7 +111,7 @@ Advertise extension support in `initialize.result.agentCapabilities._meta["sandb
Clients must feature-detect and degrade gracefully.
## 4) Recommendation for Current v2
## 4) Recommendation for Current v1
Recommended implementation order:

48
research/acp/friction.md
View file

@ -21,8 +21,8 @@ Update this file continuously during the migration.
- Date: 2026-02-10
- Area: Agent process availability
- Issue: Amp does not have a confirmed official ACP agent process in current ACP docs/research.
- Impact: Blocks full parity if Amp is required in v2 launch scope.
- Proposed direction: Treat Amp as conditional for v2.0 and support via pinned fallback only if agent process source is validated.
- Impact: Blocks full parity if Amp is required in v1 launch scope.
- Proposed direction: Treat Amp as conditional for v1.0 and support via pinned fallback only if agent process source is validated.
- Decision: Open.
- Owner: Unassigned.
- Status: open
@ -30,7 +30,7 @@ Update this file continuously during the migration.
- Date: 2026-02-10
- Area: Transport
- Issue: ACP streamable HTTP is still draft upstream; v2 requires ACP over HTTP now.
- Issue: ACP streamable HTTP is still draft upstream; v1 requires ACP over HTTP now.
- Impact: Potential divergence from upstream HTTP semantics.
- Proposed direction: Use strict JSON-RPC mapping and keep transport shim minimal/documented for later alignment.
- Decision: Open.
@ -72,7 +72,7 @@ Update this file continuously during the migration.
- Area: ACP over HTTP standardization
- Issue: Community is actively piloting both Streamable HTTP and WebSocket; no final single transport profile has emerged yet.
- Impact: Risk of rework if we overfit to one draft behavior that later shifts.
- Proposed direction: Lock v2 public contract to Streamable HTTP with ACP JSON-RPC payloads, keep implementation modular so WebSocket can be added later without breaking v2 API.
- Proposed direction: Lock v1 public contract to Streamable HTTP with ACP JSON-RPC payloads, keep implementation modular so WebSocket can be added later without breaking v1 API.
- Decision: Accepted.
- Owner: Unassigned.
- Status: in_progress
@ -121,7 +121,7 @@ Update this file continuously during the migration.
- Date: 2026-02-10
- Area: TypeScript ACP-over-HTTP client support
- Issue: Official ACP client SDK does not currently provide the exact Streamable HTTP transport behavior required by this project.
- Impact: SDK cannot target `/v2/rpc` without additional transport implementation.
- Impact: SDK cannot target `/v1/rpc` without additional transport implementation.
- Proposed direction: Embed upstream ACP SDK types/lifecycle and implement a project transport agent process for ACP-over-HTTP.
- Decision: Accepted.
- Owner: Unassigned.
@ -156,7 +156,7 @@ Update this file continuously during the migration.
- Decision: Accepted and implemented.
- Owner: Unassigned.
- Status: resolved
- Links: `server/packages/agent-management/src/agents.rs`, `server/packages/sandbox-agent/tests/v2_api.rs`
- Links: `server/packages/agent-management/src/agents.rs`, `server/packages/sandbox-agent/tests/v1_api.rs`
- Date: 2026-02-10
- Area: Inspector E2E tooling
@ -179,9 +179,9 @@ Update this file continuously during the migration.
- Links: `research/acp/todo.md`
- Date: 2026-02-10
- Area: Inspector v1-to-v2 compatibility
- Issue: Restored inspector UI expects legacy `/v1` session/event contracts that no longer exist in ACP-native v2.
- Impact: Full parity would block migration; inspector would otherwise fail to run against v2.
- Area: Inspector v1-to-v1 compatibility
- Issue: Restored inspector UI expects legacy `/v1` session/event contracts that no longer exist in ACP-native v1.
- Impact: Full parity would block migration; inspector would otherwise fail to run against v1.
- Proposed direction: Keep the restored UI and bridge to ACP with a thin compatibility client (`src/lib/legacyClient.ts`), stubbing non-parity features with explicit `TDOO` markers.
- Decision: Accepted.
- Owner: Unassigned.
@ -196,7 +196,7 @@ Update this file continuously during the migration.
- Decision: Accepted and implemented.
- Owner: Unassigned.
- Status: resolved
- Links: `server/packages/sandbox-agent/src/acp_runtime/mod.rs`, `server/packages/sandbox-agent/src/acp_runtime/mock.rs`, `server/packages/sandbox-agent/tests/v2_api.rs`, `server/packages/sandbox-agent/tests/v2_agent_process_matrix.rs`
- Links: `server/packages/sandbox-agent/src/acp_runtime/mod.rs`, `server/packages/sandbox-agent/src/acp_runtime/mock.rs`, `server/packages/sandbox-agent/tests/v1_api.rs`, `server/packages/sandbox-agent/tests/v1_agent_process_matrix.rs`
- Date: 2026-02-10
- Area: TypeScript SDK package split and ACP lifecycle
@ -210,13 +210,13 @@ Update this file continuously during the migration.
- Date: 2026-02-10
- Area: Streamable HTTP transport contract
- Issue: Ambiguity over whether `/v2/rpc` should track MCP transport negotiation (`POST` accepting SSE responses, multi-stream fanout) versus Sandbox Agent's simpler JSON-only POST contract.
- Issue: Ambiguity over whether `/v1/rpc` should track MCP transport negotiation (`POST` accepting SSE responses, multi-stream fanout) versus Sandbox Agent's simpler JSON-only POST contract.
- Impact: Without an explicit contract, clients can assume incompatible Accept/media semantics and open duplicate GET streams that receive duplicate events.
- Proposed direction: Define Sandbox Agent transport profile explicitly: `POST /v2/rpc` is JSON-only (`Content-Type` and `Accept` for `application/json`), `GET /v2/rpc` is SSE-only (`Accept: text/event-stream`), and allow only one active SSE stream per ACP connection id.
- Proposed direction: Define Sandbox Agent transport profile explicitly: `POST /v1/rpc` is JSON-only (`Content-Type` and `Accept` for `application/json`), `GET /v1/rpc` is SSE-only (`Accept: text/event-stream`), and allow only one active SSE stream per ACP connection id.
- Decision: Accepted and implemented.
- Owner: Unassigned.
- Status: resolved
- Links: `server/packages/sandbox-agent/src/router.rs`, `server/packages/sandbox-agent/src/acp_runtime/mod.rs`, `server/packages/sandbox-agent/tests/v2_api/acp_transport.rs`, `docs/advanced/acp-http-client.mdx`
- Links: `server/packages/sandbox-agent/src/router.rs`, `server/packages/sandbox-agent/src/acp_runtime/mod.rs`, `server/packages/sandbox-agent/tests/v1_api/acp_transport.rs`, `docs/advanced/acp-http-client.mdx`
- Date: 2026-02-10
- Area: Agent selection contract for ACP bootstrap/session creation
@ -226,4 +226,24 @@ Update this file continuously during the migration.
- Decision: Accepted and implemented.
- Owner: Unassigned.
- Status: resolved
- Links: `server/packages/sandbox-agent/src/router.rs`, `server/packages/sandbox-agent/src/acp_runtime/helpers.rs`, `server/packages/sandbox-agent/src/acp_runtime/mod.rs`, `server/packages/sandbox-agent/src/acp_runtime/ext_meta.rs`, `server/packages/sandbox-agent/tests/v2_api/acp_transport.rs`
- Links: `server/packages/sandbox-agent/src/router.rs`, `server/packages/sandbox-agent/src/acp_runtime/helpers.rs`, `server/packages/sandbox-agent/src/acp_runtime/mod.rs`, `server/packages/sandbox-agent/src/acp_runtime/ext_meta.rs`, `server/packages/sandbox-agent/tests/v1_api/acp_transport.rs`
- Date: 2026-02-11
- Area: ACP server simplification
- Issue: Current `/v1/rpc` runtime includes server-managed metadata/session registry and `_sandboxagent/*` ACP extensions, while the new direction is a dumb stdio proxy keyed by client-provided ACP server id.
- Impact: Requires removing extension/metadata semantics and reshaping transport to `/v1/acp/{server_id}` with per-id subprocess lifecycle.
- Proposed direction: Replace `/v1/rpc` with `/v1/acp/{server_id}` (`POST`/`GET` SSE/`DELETE`), drop connection-id headers, keep replay by `server_id`, move non-ACP concerns to HTTP endpoints, and disable OpenCode routes.
- Decision: Accepted (spec drafted).
- Owner: Unassigned.
- Status: in_progress
- Links: `research/acp/simplify-server.md`
- Date: 2026-02-11
- Area: Directory-scoped config ownership
- Issue: MCP/skills config previously traveled with session initialization payloads; simplified server needs standalone HTTP config scoped by directory.
- Impact: Requires new HTTP APIs and clear naming for per-directory/per-entry operations without ACP extension transport.
- Proposed direction: Add directory-scoped query APIs: `/v1/config/mcp?directory=...&mcpName=...` and `/v1/config/skills?directory=...&skillName=...` (name required), using v1 payload shapes for MCP/skills config values.
- Decision: Accepted (spec updated).
- Owner: Unassigned.
- Status: in_progress
- Links: `research/acp/simplify-server.md`, `docs/mcp-config.mdx`, `docs/skills-config.mdx`

2
research/acp/inspector-unimplemented.md
View file

@ -2,7 +2,7 @@
Updated: 2026-02-10
This tracks legacy inspector behaviors that do not yet have full parity on ACP v2.
This tracks legacy inspector behaviors that do not yet have full parity on ACP v1.
1. TDOO: Session `permissionMode` preconfiguration on create is not wired in ACP inspector compatibility.
2. TDOO: Session `variant` preconfiguration on create is not wired in ACP inspector compatibility.

114
research/acp/merge-acp.md
View file

@ -1,57 +1,57 @@
# Proposal: Move Static v2 HTTP Endpoints into ACP Extensions
# Proposal: Move Static v1 HTTP Endpoints into ACP Extensions
## Goal
Keep `GET /v2/health` as the only static control endpoint, except for dedicated binary filesystem transfer endpoints.
Keep `GET /v1/health` as the only static control endpoint, except for dedicated binary filesystem transfer endpoints.
Move all other current static v2 HTTP routes to ACP JSON-RPC methods (Sandbox Agent extensions under `_sandboxagent/...`) on `/v2/rpc`.
Move all other current static v1 HTTP routes to ACP JSON-RPC methods (Sandbox Agent extensions under `_sandboxagent/...`) on `/v1/rpc`.
Retain these HTTP endpoints intentionally:
- `GET /v2/fs/file`
- `PUT /v2/fs/file`
- `POST /v2/fs/upload-batch`
- `GET /v1/fs/file`
- `PUT /v1/fs/file`
- `POST /v1/fs/upload-batch`
No implementation in this proposal. This is a migration plan.
## Current State (from `server/packages/sandbox-agent/src/router.rs`)
Static v2 endpoints today:
Static v1 endpoints today:
- `GET /v2/agents`
- `POST /v2/agents/:agent/install`
- `GET /v2/sessions`
- `GET /v2/sessions/:session_id`
- `GET /v2/fs/entries`
- `GET /v2/fs/file`
- `PUT /v2/fs/file`
- `DELETE /v2/fs/entry`
- `POST /v2/fs/mkdir`
- `POST /v2/fs/move`
- `GET /v2/fs/stat`
- `POST /v2/fs/upload-batch`
- `GET /v1/agents`
- `POST /v1/agents/:agent/install`
- `GET /v1/sessions`
- `GET /v1/sessions/:session_id`
- `GET /v1/fs/entries`
- `GET /v1/fs/file`
- `PUT /v1/fs/file`
- `DELETE /v1/fs/entry`
- `POST /v1/fs/mkdir`
- `POST /v1/fs/move`
- `GET /v1/fs/stat`
- `POST /v1/fs/upload-batch`
Non-static ACP transport endpoints (remain):
- `POST /v2/rpc`
- `GET /v2/rpc` (SSE)
- `DELETE /v2/rpc`
- `POST /v1/rpc`
- `GET /v1/rpc` (SSE)
- `DELETE /v1/rpc`
Health endpoint (remain):
- `GET /v2/health`
- `GET /v1/health`
## Proposed Target Surface
Keep:
- `GET /v2/health`
- `POST/GET/DELETE /v2/rpc`
- `GET /v2/fs/file`
- `PUT /v2/fs/file`
- `POST /v2/fs/upload-batch`
- `GET /v1/health`
- `POST/GET/DELETE /v1/rpc`
- `GET /v1/fs/file`
- `PUT /v1/fs/file`
- `POST /v1/fs/upload-batch`
Remove all other static v2 control/file routes after migration.
Remove all other static v1 control/file routes after migration.
Add ACP extension methods:
@ -68,24 +68,24 @@ Add ACP extension methods:
- `_sandboxagent/fs/stat`
- `_sandboxagent/fs/upload_batch` (parallel with HTTP)
Interpretation for clients: all agent/session operations and non-binary filesystem operations move to ACP extension calls over `/v2/rpc`. Binary file transfer has a dual surface: ACP equivalents exist in parallel, but HTTP remains the primary transport for large/streaming payloads.
Interpretation for clients: all agent/session operations and non-binary filesystem operations move to ACP extension calls over `/v1/rpc`. Binary file transfer has a dual surface: ACP equivalents exist in parallel, but HTTP remains the primary transport for large/streaming payloads.
## Endpoint-to-Method Mapping
| Existing HTTP | New ACP method | Notes |
| --- | --- | --- |
| `GET /v2/agents` | `_sandboxagent/agent/list` | Response keeps current `AgentListResponse` shape for low migration risk. |
| `POST /v2/agents/:agent/install` | `_sandboxagent/agent/install` | Params include `agent`, `reinstall`, `agentVersion`, `agentProcessVersion`. |
| `GET /v2/sessions` | `_sandboxagent/session/list` | Return current `SessionListResponse` shape (not ACP unstable list shape). |
| `GET /v2/sessions/:session_id` | `_sandboxagent/session/get` | Return current `SessionInfo` shape; error on missing session. |
| `GET /v2/fs/entries` | `_sandboxagent/fs/list_entries` | Preserve path + optional `sessionId` resolution semantics. |
| `GET /v2/fs/file` | keep HTTP + `_sandboxagent/fs/read_file` | HTTP is primary because responses may require large streaming reads; ACP variant exists for compatibility/smaller payloads. |
| `PUT /v2/fs/file` | keep HTTP + `_sandboxagent/fs/write_file` | HTTP is primary for large binary writes; ACP variant exists for compatibility/smaller payloads. |
| `DELETE /v2/fs/entry` | `_sandboxagent/fs/delete_entry` | Preserve recursive directory delete behavior. |
| `POST /v2/fs/mkdir` | `_sandboxagent/fs/mkdir` | Preserve create-dir behavior. |
| `POST /v2/fs/move` | `_sandboxagent/fs/move` | Preserve `overwrite` behavior. |
| `GET /v2/fs/stat` | `_sandboxagent/fs/stat` | Preserve `FsStat` shape. |
| `POST /v2/fs/upload-batch` | keep HTTP + `_sandboxagent/fs/upload_batch` | HTTP is primary for large tar uploads; ACP variant exists for compatibility/smaller payloads. |
| `GET /v1/agents` | `_sandboxagent/agent/list` | Response keeps current `AgentListResponse` shape for low migration risk. |
| `POST /v1/agents/:agent/install` | `_sandboxagent/agent/install` | Params include `agent`, `reinstall`, `agentVersion`, `agentProcessVersion`. |
| `GET /v1/sessions` | `_sandboxagent/session/list` | Return current `SessionListResponse` shape (not ACP unstable list shape). |
| `GET /v1/sessions/:session_id` | `_sandboxagent/session/get` | Return current `SessionInfo` shape; error on missing session. |
| `GET /v1/fs/entries` | `_sandboxagent/fs/list_entries` | Preserve path + optional `sessionId` resolution semantics. |
| `GET /v1/fs/file` | keep HTTP + `_sandboxagent/fs/read_file` | HTTP is primary because responses may require large streaming reads; ACP variant exists for compatibility/smaller payloads. |
| `PUT /v1/fs/file` | keep HTTP + `_sandboxagent/fs/write_file` | HTTP is primary for large binary writes; ACP variant exists for compatibility/smaller payloads. |
| `DELETE /v1/fs/entry` | `_sandboxagent/fs/delete_entry` | Preserve recursive directory delete behavior. |
| `POST /v1/fs/mkdir` | `_sandboxagent/fs/mkdir` | Preserve create-dir behavior. |
| `POST /v1/fs/move` | `_sandboxagent/fs/move` | Preserve `overwrite` behavior. |
| `GET /v1/fs/stat` | `_sandboxagent/fs/stat` | Preserve `FsStat` shape. |
| `POST /v1/fs/upload-batch` | keep HTTP + `_sandboxagent/fs/upload_batch` | HTTP is primary for large tar uploads; ACP variant exists for compatibility/smaller payloads. |
## ACP Contract Details
@ -99,14 +99,14 @@ Add keys for new extensions (`agentList`, `agentInstall`, `fsListEntries`, `fsSt
### Filesystem Exception (Intentional)
`GET/PUT /v2/fs/file` and `POST /v2/fs/upload-batch` stay as first-class Sandbox Agent HTTP APIs.
`GET/PUT /v1/fs/file` and `POST /v1/fs/upload-batch` stay as first-class Sandbox Agent HTTP APIs.
Reason:
- These operations are host/runtime capabilities implemented by Sandbox Agent, not agent-process behavior.
- Keeping them server-owned gives consistent behavior across agents.
- ACP envelopes are JSON-RPC payloads and are not suitable for streaming very large binary files efficiently.
- `GET /v2/fs/file` specifically needs efficient streamed responses for large reads.
- `GET /v1/fs/file` specifically needs efficient streamed responses for large reads.
ACP parity note:
@ -134,16 +134,16 @@ Required change for ACP-only behavior:
- Make ACP-backed helpers connection-scoped (same as ACP methods): they must throw `NotConnectedError` when disconnected.
- Keep direct HTTP helper calls only for:
- `getHealth()`
- `readFsFile()` (`GET /v2/fs/file`)
- `writeFsFile()` (`PUT /v2/fs/file`)
- `uploadFsBatch()` (`POST /v2/fs/upload-batch`)
- `readFsFile()` (`GET /v1/fs/file`)
- `writeFsFile()` (`PUT /v1/fs/file`)
- `uploadFsBatch()` (`POST /v1/fs/upload-batch`)
- Keep ACP variants available through low-level `extMethod(...)` for advanced/smaller-payload use cases, but do not make them the SDK default path.
Package boundary after migration:
- `acp-http-client` remains protocol-pure ACP transport and generic `extMethod`/`extNotification`.
- `sandbox-agent` remains the typed wrapper that maps convenience methods to `_sandboxagent/...` extension methods.
- No direct `/v2/agents*`, `/v2/sessions*`, or non-binary `/v2/fs/*` fetches in SDK runtime code.
- No direct `/v1/agents*`, `/v1/sessions*`, or non-binary `/v1/fs/*` fetches in SDK runtime code.
- Binary file transfer keeps direct HTTP fetches on the three endpoints listed above.
- SDK policy: prefer HTTP for `readFsFile`/`writeFsFile`/`uploadFsBatch` even if ACP extension variants exist.
@ -160,7 +160,7 @@ Integration test impact (`sdks/typescript/tests/integration.test.ts`):
## Bootstrap Model (Important)
Today, first call without `x-acp-connection-id` must be `initialize`, and requires `params._meta["sandboxagent.dev"].agent`.
Today, first call to a new ACP server id should be `initialize`, and requires `params._meta["sandboxagent.dev"].agent`.
Implication after migration:
@ -177,30 +177,30 @@ Alternative (optional): introduce a runtime-only control connection mode that do
- Reuse existing router/support mapping logic where possible to keep response parity.
- Keep binary file-transfer ACP methods in parallel with HTTP (`_sandboxagent/fs/read_file`, `_sandboxagent/fs/write_file`, `_sandboxagent/fs/upload_batch`) and route both surfaces through shared implementation code.
- Advertise new capabilities in `acp_runtime/ext_meta.rs`.
- Add ACP extension tests for each new method in `server/packages/sandbox-agent/tests/v2_api/acp_extensions.rs`.
- Add ACP extension tests for each new method in `server/packages/sandbox-agent/tests/v1_api/acp_extensions.rs`.
### Phase 2: Migrate Clients (No HTTP Route Removal Yet)
- TypeScript SDK (`sdks/typescript/src/client.ts`):
- Repoint `listAgents`, `installAgent`, `listSessions`, `getSession`, `listFsEntries`, `deleteFsEntry`, `mkdirFs`, `moveFs`, and `statFs` to ACP extension calls.
- Keep `readFsFile`, `writeFsFile`, and `uploadFsBatch` on HTTP endpoints.
- Remove direct runtime fetch usage for `/v2/agents*`, `/v2/sessions*`, and non-binary `/v2/fs/*`.
- Remove direct runtime fetch usage for `/v1/agents*`, `/v1/sessions*`, and non-binary `/v1/fs/*`.
- Keep method names stable for callers.
- Move these methods to connected-only semantics (`NotConnectedError` when disconnected).
- CLI (`server/packages/sandbox-agent/src/cli.rs`):
- Make `api agents list/install` call ACP extension methods (via ACP post flow), not direct `/v2/agents*` HTTP calls.
- Make `api agents list/install` call ACP extension methods (via ACP post flow), not direct `/v1/agents*` HTTP calls.
- Inspector flow/docs:
- Stop depending on `GET /v2/agents` in startup path; use ACP extension instead.
- Stop depending on `GET /v1/agents` in startup path; use ACP extension instead.
### Phase 3: Remove Static Endpoints (Except Health + Binary FS Transfer)
- Remove route registrations for `/v2/agents*`, `/v2/sessions*`, `/v2/fs/entries`, `/v2/fs/entry`, `/v2/fs/mkdir`, `/v2/fs/move`, `/v2/fs/stat` from `router.rs`.
- Keep `/v2/health`, `/v2/rpc`, `GET /v2/fs/file`, `PUT /v2/fs/file`, and `POST /v2/fs/upload-batch`.
- Remove route registrations for `/v1/agents*`, `/v1/sessions*`, `/v1/fs/entries`, `/v1/fs/entry`, `/v1/fs/mkdir`, `/v1/fs/move`, `/v1/fs/stat` from `router.rs`.
- Keep `/v1/health`, `/v1/rpc`, `GET /v1/fs/file`, `PUT /v1/fs/file`, and `POST /v1/fs/upload-batch`.
- Optional short deprecation period: convert removed routes to `410 Gone` with explicit extension method in `detail`.
### Phase 4: Docs/OpenAPI/Test Cleanup
- Regenerate `docs/openapi.json` (should now primarily describe `/v2/health`, `/v2/rpc`, and retained binary fs transfer endpoints).
- Regenerate `docs/openapi.json` (should now primarily describe `/v1/health`, `/v1/rpc`, and retained binary fs transfer endpoints).
- Update:
- `docs/cli.mdx`
- `docs/inspector.mdx`
@ -237,6 +237,6 @@ Inspector:
## Open Decisions
1. Should removed `/v2/agents*`, `/v2/sessions*`, and non-binary `/v2/fs/*` return `410` for one release or be dropped immediately?
1. Should removed `/v1/agents*`, `/v1/sessions*`, and non-binary `/v1/fs/*` return `410` for one release or be dropped immediately?
2. Do we keep a strict response-shape parity layer for session/file methods, or normalize to ACP-native shapes?
3. Should `/` service-root remain as informational HTTP, or be treated as out-of-scope for this “only health static + binary fs transfer” policy?

40
research/acp/migration-steps.md
View file

@ -11,7 +11,7 @@
The migration test plan is intentionally collapsed to avoid duplicate coverage.
1. ACP protocol conformance
2. Transport contract (`/v2/rpc`)
2. Transport contract (`/v1/rpc`)
3. End-to-end agent process matrix (core flow + cancel + HITL + streaming)
4. Installer suite (explicit + lazy + registry/fallback provenance)
5. Security/auth isolation
@ -56,7 +56,7 @@ Validation gate:
2. Implement agent process process manager (spawn, supervise, reconnect policy).
3. Implement JSON-RPC bridge: HTTP POST/SSE <-> agent process stdio.
4. Add connection registry keyed by `X-ACP-Connection-Id`.
5. Include unstable ACP methods in the v2 profile (`session/list`, `session/fork`, `session/resume`, `session/set_model`, `$/cancel_request`).
5. Include unstable ACP methods in the v1 profile (`session/list`, `session/fork`, `session/resume`, `session/set_model`, `$/cancel_request`).
Exit criteria:
@ -64,11 +64,11 @@ Exit criteria:
Validation gate:
- End-to-end ACP flow test over `/v2/rpc` (request/response + streamed notifications).
- End-to-end ACP flow test over `/v1/rpc` (request/response + streamed notifications).
- Cancellation test (`session/cancel`) with proper terminal response behavior.
- HITL request/response round-trip test (`session/request_permission` path).
- SSE ordering and reconnection behavior test (`Last-Event-ID` replay path).
- Explicit close test (`DELETE /v2/rpc`) including idempotent double-close behavior.
- Explicit close test (`DELETE /v1/rpc`) including idempotent double-close behavior.
- Unstable ACP methods validation (`session/list`, `session/fork`, `session/resume`, `session/set_model`, `$/cancel_request`) for agent processes that advertise support.
## Phase 3: Installer Refactor
@ -79,7 +79,7 @@ Validation gate:
4. Add install verification command per agent process.
5. Add ACP registry integration for install metadata + fallback sources.
6. Generate install instructions from manifest and expose provenance (`registry` or `fallback`) in API/CLI.
7. Implement lazy install path on first `/v2/rpc` initialize (with per-agent install lock and idempotent results).
7. Implement lazy install path on first `/v1/rpc` initialize (with per-agent install lock and idempotent results).
8. Add config to disable lazy install for preprovisioned environments.
Exit criteria:
@ -92,39 +92,39 @@ Validation gate:
- Lazy install on first ACP `initialize` test.
- Reinstall/version/provenance assertions.
## Phase 4: v2 HTTP API
## Phase 4: v1 HTTP API
1. Mount `/v2/rpc` POST and SSE endpoints.
2. Add `/v2/health`, `/v2/agents`, `/v2/agents/{agent}/install`.
1. Mount `/v1/rpc` POST and SSE endpoints.
2. Add `/v1/health`, `/v1/agents`, `/v1/agents/{agent}/install`.
3. Add auth integration on connection lifecycle.
4. Keep `/ui/` inspector route and migrate inspector backend calls to ACP v2 transport.
4. Keep `/ui/` inspector route and migrate inspector backend calls to ACP v1 transport.
5. Remove v1 OpenAPI generation from default docs build.
Exit criteria:
- v2 endpoints documented and passing integration tests.
- v1 endpoints documented and passing integration tests.
Validation gate:
- Contract tests for all `/v2` endpoints (`/v2/rpc`, `/v2/health`, `/v2/agents`, install).
- Contract tests for all `/v1` endpoints (`/v1/rpc`, `/v1/health`, `/v1/agents`, install).
- Auth tests (valid, missing, invalid token).
- Error mapping tests (bad envelope, unknown connection, timeout paths).
- `/v1/*` removal contract test (HTTP 410 + stable payload).
- Inspector ACP `agent-browser` flow tests pass.
- `DELETE /v2/rpc` close contract tests pass.
- `DELETE /v1/rpc` close contract tests pass.
## Phase 5: SDK and CLI v2
## Phase 5: SDK and CLI v1
1. Add ACP transport client in `sdks/typescript` by embedding `@agentclientprotocol/sdk` (no in-house ACP reimplementation).
2. Implement custom ACP-over-HTTP transport agent process in our SDK (official ACP client SDK does not provide required Streamable HTTP behavior out of the box).
3. Add inspector frontend client wiring to use ACP-over-HTTP transport primitives.
4. Add CLI commands for sending raw ACP envelopes and streaming ACP messages.
5. Remove v1-only SDK/CLI methods (or hard-fail with "v1 removed").
6. Regenerate docs to v2 ACP contract.
6. Regenerate docs to v1 ACP contract.
Exit criteria:
- SDK can complete a full ACP prompt turn over `/v2/rpc`.
- SDK can complete a full ACP prompt turn over `/v1/rpc`.
Validation gate:
@ -136,9 +136,9 @@ Validation gate:
1. Replace v1 HTTP/session tests with ACP transport contract tests.
2. Add smoke tests per supported agent process.
Current deterministic matrix: `server/packages/sandbox-agent/tests/v2_agent_process_matrix.rs`.
Current deterministic matrix: `server/packages/sandbox-agent/tests/v1_agent_process_matrix.rs`.
3. Add canary rollout notes directly in `docs/quickstart.mdx`, `docs/cli.mdx`, and `docs/sdks/typescript.mdx`.
4. Update docs for v2 ACP, `/v1/*` removal, inspector ACP behavior, and SDK usage.
4. Update docs for v1 ACP, `/v1/*` removal, inspector ACP behavior, and SDK usage.
5. Keep v1 endpoints hard-removed (`410`) until/unless a separate compatibility project is approved.
Exit criteria:
@ -155,9 +155,9 @@ Validation gate:
## Phase 7: OpenCode <-> ACP Bridge (Dedicated Step)
1. Keep `/opencode/*` commented out/disabled through Phases 1-6.
2. Implement OpenCode <-> ACP bridge on top of v2 ACP runtime.
2. Implement OpenCode <-> ACP bridge on top of v1 ACP runtime.
3. Re-enable `server/packages/sandbox-agent/src/opencode_compat.rs` routes/tests at full capability.
4. Add dedicated integration tests that validate OpenCode SDK/TUI flows through ACP v2 internals.
4. Add dedicated integration tests that validate OpenCode SDK/TUI flows through ACP v1 internals.
Exit criteria:
@ -170,5 +170,5 @@ Validation gate:
## Compatibility Layer (optional future project)
1. No compatibility layer is in the current v2 scope.
1. No compatibility layer is in the current v1 scope.
2. If later approved, it should be a separate project with a dedicated spec and test matrix.

4
research/acp/missing-features-spec/01-questions.md
View file

@ -4,9 +4,9 @@
## Summary
v1 had a full question subsystem: agent requests a question from the user, client replies with an answer or rejection, and the system tracks question status. v2 has partial stub implementation in mock only.
v1 had a full question subsystem: agent requests a question from the user, client replies with an answer or rejection, and the system tracks question status. v1 has partial stub implementation in mock only.
## Current v2 State
## Current v1 State
- `_sandboxagent/session/request_question` is declared as a constant in `acp_runtime/mod.rs:33`
- Advertised in capability injection (`extensions.sessionRequestQuestion: true`)

30
research/acp/missing-features-spec/04-filesystem-api.md
View file

@ -4,13 +4,13 @@
## Summary
v1 had 8 filesystem endpoints. v2 has only ACP `fs/read_text_file` + `fs/write_text_file` (text-only, agent->client direction). The full filesystem API should be re-implemented as Sandbox Agent-specific HTTP contracts at `/v2/fs/*`.
v1 had 8 filesystem endpoints. v1 has only ACP `fs/read_text_file` + `fs/write_text_file` (text-only, agent->client direction). The full filesystem API should be re-implemented as Sandbox Agent-specific HTTP contracts at `/v1/fs/*`.
## Current v2 State
## Current v1 State
- ACP stable: `fs/read_text_file`, `fs/write_text_file` (client methods invoked by agents, text-only)
- No HTTP filesystem endpoints exist in current `router.rs`
- `rfds-vs-extensions.md` confirms: "Already extension (`/v2/fs/*` custom HTTP surface)"
- `rfds-vs-extensions.md` confirms: "Already extension (`/v1/fs/*` custom HTTP surface)"
- CLAUDE.md: "Filesystem and terminal APIs remain Sandbox Agent-specific HTTP contracts and are not ACP"
## v1 Reference (source commit)
@ -356,32 +356,32 @@ async fn fs_upload_batch(
## Implementation Plan
### New v2 Endpoints
### New v1 Endpoints
| Method | Path | Description |
|--------|------|-------------|
| GET | `/v2/fs/entries` | List directory entries |
| GET | `/v2/fs/file` | Read file raw bytes |
| PUT | `/v2/fs/file` | Write file raw bytes |
| DELETE | `/v2/fs/entry` | Delete file or directory |
| POST | `/v2/fs/mkdir` | Create directory |
| POST | `/v2/fs/move` | Move/rename |
| GET | `/v2/fs/stat` | File metadata |
| POST | `/v2/fs/upload-batch` | Upload tar archive |
| GET | `/v1/fs/entries` | List directory entries |
| GET | `/v1/fs/file` | Read file raw bytes |
| PUT | `/v1/fs/file` | Write file raw bytes |
| DELETE | `/v1/fs/entry` | Delete file or directory |
| POST | `/v1/fs/mkdir` | Create directory |
| POST | `/v1/fs/move` | Move/rename |
| GET | `/v1/fs/stat` | File metadata |
| POST | `/v1/fs/upload-batch` | Upload tar archive |
### Files to Modify
| File | Change |
|------|--------|
| `server/packages/sandbox-agent/src/router.rs` | Add all 8 `/v2/fs/*` endpoints with handlers (port from v1 with v2 path prefix) |
| `server/packages/sandbox-agent/src/router.rs` | Add all 8 `/v1/fs/*` endpoints with handlers (port from v1 with v1 path prefix) |
| `server/packages/sandbox-agent/src/cli.rs` | Add CLI `fs` subcommands (list, read, write, delete, mkdir, move, stat) |
| `sdks/typescript/src/client.ts` | Add filesystem methods to SDK |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Add filesystem endpoint tests |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Add filesystem endpoint tests |
### Docs to Update
| Doc | Change |
|-----|--------|
| `docs/openapi.json` | Add `/v2/fs/*` endpoint specs |
| `docs/openapi.json` | Add `/v1/fs/*` endpoint specs |
| `docs/cli.mdx` | Add `fs` subcommand documentation |
| `docs/sdks/typescript.mdx` | Document filesystem SDK methods |

18
research/acp/missing-features-spec/05-health-endpoint.md
View file

@ -1,12 +1,12 @@
# Feature 5: Health Endpoint
**Implementation approach:** Enhance existing `GET /v2/health`
**Implementation approach:** Enhance existing `GET /v1/health`
## Summary
v1 had a typed `HealthResponse` with detailed status. v2 `GET /v2/health` exists but returns only `{ status: "ok", api_version: "v2" }`. Needs enrichment.
v1 had a typed `HealthResponse` with detailed status. v1 `GET /v1/health` exists but returns only `{ status: "ok", api_version: "v1" }`. Needs enrichment.
## Current v2 State
## Current v1 State
From `router.rs:332-346`:
@ -17,10 +17,10 @@ pub struct HealthResponse {
pub api_version: String,
}
async fn get_v2_health() -> Json<HealthResponse> {
async fn get_v1_health() -> Json<HealthResponse> {
Json(HealthResponse {
status: "ok".to_string(),
api_version: "v2".to_string(),
api_version: "v1".to_string(),
})
}
```
@ -72,19 +72,19 @@ pub struct HealthResponse {
}
```
`GET /v2/health` should mirror v1 semantics and response shape (ported from commit `8ecd27bc24e62505d7aa4c50cbdd1c9dbb09f836`), while keeping the v2 route path.
`GET /v1/health` should mirror v1 semantics and response shape (ported from commit `8ecd27bc24e62505d7aa4c50cbdd1c9dbb09f836`), while keeping the v1 route path.
### Files to Modify
| File | Change |
|------|--------|
| `server/packages/sandbox-agent/src/router.rs` | Port v1 health response types/logic onto `GET /v2/health` |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Update health endpoint test for full v1-parity payload |
| `server/packages/sandbox-agent/src/router.rs` | Port v1 health response types/logic onto `GET /v1/health` |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Update health endpoint test for full v1-parity payload |
| `sdks/typescript/src/client.ts` | Update `HealthResponse` type |
### Docs to Update
| Doc | Change |
|-----|--------|
| `docs/openapi.json` | Update `/v2/health` response schema |
| `docs/openapi.json` | Update `/v1/health` response schema |
| `docs/sdks/typescript.mdx` | Document enriched health response |

14
research/acp/missing-features-spec/06-server-status.md
View file

@ -1,14 +1,14 @@
# Feature 6: Server Status
**Implementation approach:** Extension fields on `GET /v2/agents` and `GET /v2/health`
**Implementation approach:** Extension fields on `GET /v1/agents` and `GET /v1/health`
## Summary
v1 had `ServerStatus` (Running/Stopped/Error) and `ServerStatusInfo` (baseUrl, lastError, restartCount, uptimeMs) per agent. v2 has none of this. Add server/agent process status tracking.
v1 had `ServerStatus` (Running/Stopped/Error) and `ServerStatusInfo` (baseUrl, lastError, restartCount, uptimeMs) per agent. v1 has none of this. Add server/agent process status tracking.
## Current v2 State
## Current v1 State
`GET /v2/agents` returns `AgentInfo` with install state only:
`GET /v1/agents` returns `AgentInfo` with install state only:
```rust
pub struct AgentInfo {
@ -132,13 +132,13 @@ Only include `server_status` for agents that use shared processes (Codex, OpenCo
| File | Change |
|------|--------|
| `server/packages/sandbox-agent/src/acp_runtime/mod.rs` | Track agent process lifecycle (start/stop/error/restart count) per `AgentId`; expose `status_snapshot()` method |
| `server/packages/sandbox-agent/src/router.rs` | Add `ServerStatus`, `ServerStatusInfo` types; add `server_status` to `AgentInfo`; query runtime for status in `get_v2_agents` |
| `server/packages/sandbox-agent/src/router.rs` | Add `ServerStatus`, `ServerStatusInfo` types; add `server_status` to `AgentInfo`; query runtime for status in `get_v1_agents` |
| `sdks/typescript/src/client.ts` | Update `AgentInfo` type with `serverStatus` |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Test server status in agent listing |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Test server status in agent listing |
### Docs to Update
| Doc | Change |
|-----|--------|
| `docs/openapi.json` | Update `/v2/agents` response with `server_status` |
| `docs/openapi.json` | Update `/v1/agents` response with `server_status` |
| `docs/sdks/typescript.mdx` | Document `serverStatus` field |

8
research/acp/missing-features-spec/07-session-termination.md
View file

@ -4,12 +4,12 @@
## Summary
v1 had explicit session termination (`POST /v1/sessions/{id}/terminate`). v2 only has `session/cancel` (turn cancellation, not session kill) and `DELETE /v2/rpc` (connection close, not session termination). Need explicit session destroy/terminate semantics.
v1 had explicit session termination (`POST /v1/sessions/{id}/terminate`). v1 only has `session/cancel` (turn cancellation, not session kill) and `DELETE /v1/rpc` (connection close, not session termination). Need explicit session destroy/terminate semantics.
## Current v2 State
## Current v1 State
- `session/cancel` — cancels an in-flight prompt turn only
- `DELETE /v2/rpc` — closes the HTTP connection, does **not** terminate the session
- `DELETE /v1/rpc` — closes the HTTP connection, does **not** terminate the session
- `_sandboxagent/session/detach` — detaches a session from a connection (multi-client visibility)
- No session termination/deletion exists
- `rfds-vs-extensions.md`: "Session Termination: Not covered by ACP. Only implement if product explicitly requires termination semantics beyond session/cancel"
@ -112,7 +112,7 @@ Response:
| `server/packages/sandbox-agent/src/acp_runtime/mod.rs` | Add `_sandboxagent/session/terminate` handler; add session removal from registry; add process kill logic |
| `server/packages/sandbox-agent/src/acp_runtime/mock.rs` | Add mock terminate support |
| `sdks/typescript/src/client.ts` | Add `terminateSession(sessionId)` method |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Add session termination test |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Add session termination test |
### Docs to Update

6
research/acp/missing-features-spec/08-model-variants.md
View file

@ -6,9 +6,9 @@
## Summary
v1 had `AgentModelInfo.variants`, `AgentModelInfo.defaultVariant`, and `CreateSessionRequest.variant`. v2 already has `_sandboxagent/session/list_models` but the variant fields need to be verified and the session-creation variant selection needs to work end-to-end.
v1 had `AgentModelInfo.variants`, `AgentModelInfo.defaultVariant`, and `CreateSessionRequest.variant`. v1 already has `_sandboxagent/session/list_models` but the variant fields need to be verified and the session-creation variant selection needs to work end-to-end.
## Current v2 State
## Current v1 State
From `acp_runtime/mod.rs`, `_sandboxagent/session/list_models` is implemented and returns:
- `availableModels[]` with `modelId`, `name`, `description`
@ -120,7 +120,7 @@ The runtime should forward this variant to the agent process (e.g., as a model p
| `server/packages/sandbox-agent/src/acp_runtime/mod.rs` | Verify `list_models` response includes `variants`/`defaultVariant`; extract and forward `variant` from `session/new` `_meta` |
| `server/packages/sandbox-agent/src/acp_runtime/mock.rs` | Add variant support to mock model listing |
| `sdks/typescript/src/client.ts` | Update `listModels` return type to include variants |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Add model variants test |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Add model variants test |
### Docs to Update

6
research/acp/missing-features-spec/10-include-raw.md
View file

@ -6,9 +6,9 @@
## Summary
v1 had an `include_raw` option that preserved the original agent JSON alongside normalized events. The `UniversalEvent.raw` field held the verbatim agent output. v2 has `_sandboxagent/agent/unparsed` for parse errors but no mechanism for clients to request raw agent payloads alongside normalized ACP events.
v1 had an `include_raw` option that preserved the original agent JSON alongside normalized events. The `UniversalEvent.raw` field held the verbatim agent output. v1 has `_sandboxagent/agent/unparsed` for parse errors but no mechanism for clients to request raw agent payloads alongside normalized ACP events.
## Current v2 State
## Current v1 State
- `_sandboxagent/agent/unparsed` — sends notifications when the runtime fails to parse agent output (error recovery only)
- No option for clients to request raw agent JSON alongside normal ACP events
@ -43,7 +43,7 @@ When `include_raw=true`, each `UniversalEvent` included the verbatim JSON the ag
### Extension Design
Since v2 agents speak ACP natively (JSON-RPC), the "raw" concept changes:
Since v1 agents speak ACP natively (JSON-RPC), the "raw" concept changes:
- For ACP-native agents: raw = the ACP JSON-RPC envelope itself (which clients already see)
- For non-native agents or runtime-synthesized events: raw = the original agent output before transformation

24
research/acp/missing-features-spec/12-agent-listing.md
View file

@ -1,14 +1,14 @@
# Feature 12: Agent Listing (Typed Response)
**Implementation approach:** Enhance existing `GET /v2/agents`
**Implementation approach:** Enhance existing `GET /v1/agents`
## Summary
v1 `GET /v1/agents` returned a typed `AgentListResponse` with `installed`, `credentialsAvailable`, `path`, `capabilities`, `serverStatus`. v2 `GET /v2/agents` returns a basic `AgentInfo` with only install state. Needs enrichment.
v1 `GET /v1/agents` returned a typed `AgentListResponse` with `installed`, `credentialsAvailable`, `path`, `capabilities`, `serverStatus`. v1 `GET /v1/agents` returns a basic `AgentInfo` with only install state. Needs enrichment.
This feature also carries pre-session models/modes as optional fields when the agent is installed (Feature #13), rather than using separate model/mode endpoints.
## Current v2 State
## Current v1 State
From `router.rs:265-275`:
@ -183,19 +183,19 @@ fn agent_capabilities_for(agent: AgentId) -> AgentCapabilities {
### Enriched AgentInfo
Merge v2 install fields with v1 richness:
Merge v1 install fields with v1 richness:
```rust
pub struct AgentInfo {
pub id: String,
pub installed: bool, // convenience: is fully installed
pub credentials_available: bool, // from credential extraction
pub native_required: bool, // keep from v2
pub native_installed: bool, // keep from v2
pub native_version: Option<String>, // keep from v2
pub agent_process_installed: bool, // keep from v2
pub agent_process_source: Option<String>, // keep from v2
pub agent_process_version: Option<String>, // keep from v2
pub native_required: bool, // keep from v1
pub native_installed: bool, // keep from v1
pub native_version: Option<String>, // keep from v1
pub agent_process_installed: bool, // keep from v1
pub agent_process_source: Option<String>, // keep from v1
pub agent_process_version: Option<String>, // keep from v1
pub path: Option<String>, // from resolve_binary()
pub capabilities: AgentCapabilities, // full v1 capability set
pub server_status: Option<AgentServerStatus>, // from Feature #6
@ -212,11 +212,11 @@ pub struct AgentInfo {
| `server/packages/sandbox-agent/src/router.rs` | Enrich `AgentInfo` and `AgentCapabilities` structs; add `agent_capabilities_for()` static mapping; add credential check; add convenience `installed` field; add optional `models`/`modes` for installed agents |
| `server/packages/agent-management/src/agents.rs` | Expose credential availability check and `resolve_binary()` if not already present |
| `sdks/typescript/src/client.ts` | Update `AgentInfo` and `AgentCapabilities` types |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Update agent listing test assertions |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Update agent listing test assertions |
### Docs to Update
| Doc | Change |
|-----|--------|
| `docs/openapi.json` | Update `/v2/agents` response schema with full `AgentCapabilities` |
| `docs/openapi.json` | Update `/v1/agents` response schema with full `AgentCapabilities` |
| `docs/sdks/typescript.mdx` | Document enriched agent listing |

10
research/acp/missing-features-spec/13-models-modes-listing.md
View file

@ -4,12 +4,12 @@
## Summary
v1 exposed pre-session model/mode discovery via separate endpoints. For v2, models and modes should be optional fields on the agent response payload (only when the agent is installed), with lazy population for dynamic agents.
v1 exposed pre-session model/mode discovery via separate endpoints. For v1, models and modes should be optional fields on the agent response payload (only when the agent is installed), with lazy population for dynamic agents.
## Current v2 State
## Current v1 State
- `_sandboxagent/session/list_models` works but requires an active ACP connection and session
- `GET /v2/agents` does not include pre-session model/mode metadata
- `GET /v1/agents` does not include pre-session model/mode metadata
- v1 had static per-agent mode definitions (`agent_modes_for()` in `router.rs`)
- v1 had dynamic model fetching (Claude/Codex/OpenCode), plus static model lists for Amp/Mock
@ -57,11 +57,11 @@ Model variants are explicitly out of scope for this implementation pass.
| `server/packages/sandbox-agent/src/router.rs` | Enrich agent response type/handlers to optionally include models + modes |
| `server/packages/sandbox-agent/src/acp_runtime/mod.rs` | Expose model query support for control-plane enrichment without requiring an active session |
| `sdks/typescript/src/client.ts` | Extend `AgentInfo` type with optional `models`, `defaultModel`, `modes` |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Add assertions for installed vs non-installed agent response shapes |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Add assertions for installed vs non-installed agent response shapes |
## Docs to Update
| Doc | Change |
|-----|--------|
| `docs/openapi.json` | Update `/v2/agents` (and agent detail endpoint if present) schema with optional `models`/`modes` |
| `docs/openapi.json` | Update `/v1/agents` (and agent detail endpoint if present) schema with optional `models`/`modes` |
| `docs/sdks/typescript.mdx` | Document optional model/mode fields on agent response |

6
research/acp/missing-features-spec/14-message-attachments.md
View file

@ -4,9 +4,9 @@
## Summary
v1 `MessageRequest.attachments` allowed sending file attachments (path, mime, filename) with prompts. v2 ACP `embeddedContext` is only partial. Need to support file attachments in prompt messages.
v1 `MessageRequest.attachments` allowed sending file attachments (path, mime, filename) with prompts. v1 ACP `embeddedContext` is only partial. Need to support file attachments in prompt messages.
## Current v2 State
## Current v1 State
- ACP `session/prompt` accepts `params.content` as the prompt text
- No attachment mechanism in the current ACP prompt flow
@ -122,7 +122,7 @@ The runtime extracts attachments from `_meta` and transforms them per agent:
| `server/packages/sandbox-agent/src/acp_runtime/mod.rs` | Extract `attachments` from `session/prompt` `_meta`; transform per agent before forwarding |
| `server/packages/sandbox-agent/src/acp_runtime/mock.rs` | Add mock handling for attachments |
| `sdks/typescript/src/client.ts` | Add `attachments` option to prompt method |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Add attachment prompt test |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Add attachment prompt test |
### Docs to Update

6
research/acp/missing-features-spec/15-session-creation-richness.md
View file

@ -4,13 +4,13 @@
## Summary
v1 `CreateSessionRequest` had `mcp` (full MCP server config with OAuth, env headers, bearer tokens), `skills` (sources with git refs), `agent_version`, `directory`. v2 needs to support these at session creation time.
v1 `CreateSessionRequest` had `mcp` (full MCP server config with OAuth, env headers, bearer tokens), `skills` (sources with git refs), `agent_version`, `directory`. v1 needs to support these at session creation time.
## Current v2 State — MOSTLY IMPLEMENTED
## Current v1 State — MOSTLY IMPLEMENTED
Investigation shows that **most of these are already supported** via `_meta.sandboxagent.dev` passthrough in `session/new`:
| Field | v1 | v2 Status | v2 Mechanism |
| Field | v1 | v1 Status | v1 Mechanism |
|-------|-----|-----------|-------------|
| `directory` | `CreateSessionRequest.directory` | **Implemented** | `cwd` parameter extracted from payload |
| `agent_version` | `CreateSessionRequest.agent_version` | **Implemented** | `_meta.sandboxagent.dev.agentVersionRequested` (stored, forwarded) |

18
research/acp/missing-features-spec/16-session-info.md
View file

@ -1,12 +1,12 @@
# Feature 16: Session Info
**Implementation approach:** New HTTP endpoints (`GET /v2/sessions`, `GET /v2/sessions/{id}`)
**Implementation approach:** New HTTP endpoints (`GET /v1/sessions`, `GET /v1/sessions/{id}`)
## Summary
v1 `SessionInfo` tracked `event_count`, `created_at`, `updated_at`, and full `mcp` config. v2 has session data in the ACP runtime's `MetaSession` struct but no HTTP endpoints to query it. Add REST endpoints for session listing and detail.
v1 `SessionInfo` tracked `event_count`, `created_at`, `updated_at`, and full `mcp` config. v1 has session data in the ACP runtime's `MetaSession` struct but no HTTP endpoints to query it. Add REST endpoints for session listing and detail.
## Current v2 State
## Current v1 State
### Internal Session Tracking
@ -117,15 +117,15 @@ fn build_session_info(state: &SessionState) -> SessionInfo {
### New HTTP Endpoints
```
GET /v2/sessions -> SessionListResponse
GET /v2/sessions/{id} -> SessionInfo
GET /v1/sessions -> SessionListResponse
GET /v1/sessions/{id} -> SessionInfo
```
These are control-plane HTTP endpoints (not ACP), providing session visibility without requiring an active ACP connection.
### Response Types
The v2 `SessionInfo` should be a superset of v1 fields, adapted for ACP:
The v1 `SessionInfo` should be a superset of v1 fields, adapted for ACP:
```rust
#[derive(Debug, Serialize, JsonSchema, ToSchema)]
@ -156,15 +156,15 @@ Need to add:
| File | Change |
|------|--------|
| `server/packages/sandbox-agent/src/router.rs` | Add `GET /v2/sessions` and `GET /v2/sessions/{id}` handlers; add response types |
| `server/packages/sandbox-agent/src/router.rs` | Add `GET /v1/sessions` and `GET /v1/sessions/{id}` handlers; add response types |
| `server/packages/sandbox-agent/src/acp_runtime/mod.rs` | Add `created_at` to `MetaSession`; add `ended` tracking; expose `list_sessions()` and `get_session()` public methods |
| `sdks/typescript/src/client.ts` | Add `listSessions()` and `getSession(id)` methods |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Add session listing and detail tests |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Add session listing and detail tests |
### Docs to Update
| Doc | Change |
|-----|--------|
| `docs/openapi.json` | Add `/v2/sessions` and `/v2/sessions/{id}` endpoint specs |
| `docs/openapi.json` | Add `/v1/sessions` and `/v1/sessions/{id}` endpoint specs |
| `docs/cli.mdx` | Add CLI `sessions list` and `sessions info` commands |
| `docs/sdks/typescript.mdx` | Document session listing SDK methods |

8
research/acp/missing-features-spec/17-error-termination-metadata.md
View file

@ -4,9 +4,9 @@
## Summary
v1 captured `exit_code`, structured `StderrOutput` (head/tail/truncated) when a session ended due to error. v2 loses this metadata. Need to capture and expose process termination details.
v1 captured `exit_code`, structured `StderrOutput` (head/tail/truncated) when a session ended due to error. v1 loses this metadata. Need to capture and expose process termination details.
## Current v2 State
## Current v1 State
- Agent process lifecycle is managed in `acp_runtime/mod.rs`
- Process exit is detected but error metadata (exit code, stderr) is not captured or forwarded
@ -171,7 +171,7 @@ When an agent process terminates with an error:
### Session Info Integration
Termination metadata should be accessible via:
- `GET /v2/sessions/{id}` (Feature #16) — include `terminationInfo` in response when session has ended
- `GET /v1/sessions/{id}` (Feature #16) — include `terminationInfo` in response when session has ended
- `session/list` ACP response — include termination status in session entries
### Files to Modify
@ -181,7 +181,7 @@ Termination metadata should be accessible via:
| `server/packages/sandbox-agent/src/acp_runtime/mod.rs` | Add stderr capture (head/tail buffer) on agent process; capture exit code; emit `_sandboxagent/session/ended`; store v1-shaped termination info in `MetaSession` |
| `server/packages/sandbox-agent/src/acp_runtime/mock.rs` | Add mock error termination scenario (e.g., when prompt contains "crash") |
| `sdks/typescript/src/client.ts` | Add `TerminationInfo` type; expose on session events and session info |
| `server/packages/sandbox-agent/tests/v2_api.rs` | Add error termination metadata test |
| `server/packages/sandbox-agent/tests/v1_api.rs` | Add error termination metadata test |
### Docs to Update

Some files were not shown because too many files have changed in this diff Show more