agent_process_status() for mock always returned Some(...) even when the
launcher file did not exist. This caused install_agent_process() to
short-circuit with "already installed" and never write the launcher
script. Fix by checking that the launcher file exists before reporting
the mock agent as installed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add hidden `mock-agent-process` CLI subcommand implementing a stdio
JSON-RPC echo agent (ported from examples/mock-acp-agent)
- Update write_mock_agent_process_launcher() to exec the new subcommand
instead of exiting with error
- Update sdks/react to support React 18 and 19 peer dependencies
- Update @types/react to v19 across workspace (pnpm override + inspector)
- Fix RefObject<T | null> compatibility for React 19 useRef() signatures
- Add version reference replacement logic to release update_version.ts
covering all docs, examples, and code files listed in CLAUDE.md
- Add missing files to CLAUDE.md Install Version References list
(architecture.mdx, boxlite, modal, computesdk docs and examples)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Bump @agentclientprotocol/sdk from 0.14.1 to 0.16.1 in acp-http-client
- Update adapters.json to reflect new SDK version
- Migrate unstableListSessions to listSessions (stabilized in SDK 0.16.0)
- Add CLAUDE.md guidance: request token location before e2e agent testing
All 5 ACP adapters remain at their latest versions. E2E testing confirms
Claude, Codex, Pi, and Cursor agents work end-to-end with credentials.
Co-authored-by: Claude Haiku 4.5 <noreply@anthropic.com>
Add comprehensive tracing instrumentation across the entire agent startup path (gigacode CLI, ACP HTTP adapter, agent installation, and process spawning) to enable detailed performance profiling. Replace npm-based agent process launchers that use npx (incurring resolution overhead on every spawn) with pre-installed npm packages, reducing startup latency. Improve error diagnostics when agent processes crash by capturing exit codes and stderr tails. Update error handling to map exited processes to dedicated error variants with actionable error messages.
Co-authored-by: Claude Haiku 4.5 <noreply@anthropic.com>
* feat: add configuration for model, mode, and thought level
* docs: document Claude effort-level filesystem config
* fix: prevent panic on empty modes/thoughtLevels in parse_agent_config
Use `.first()` with safe fallback instead of direct `[0]` index access,
which would panic if the Vec is empty and no default is set.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: harden session lifecycle and align cli.mdx example with claude.json
- destroySession: wrap session/cancel RPC in try/catch so local cleanup
always succeeds even when the agent is unreachable
- createSession/resumeOrCreateSession: clean up the remote session if
post-creation config calls (setMode/setModel/setThoughtLevel) fail,
preventing leaked orphan sessions
- cli.mdx: fix example output to match current claude.json (model name,
model order, and populated modes)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: harden session lifecycle and align config persistence logic
- resumeOrCreateSession: Remove destroy-on-error for the resume path. Config
errors now propagate without destroying a pre-existing session. The destroy
pattern remains in createSession (where the session is newly created and has
no prior state to preserve).
- setSessionMode fallback: When session/set_mode returns -32601 and the
fallback uses session/set_config_option, now keep modes.currentModeId
in sync with the updated currentValue. Prevents stale cached state in
getModes() when the fallback path is used.
- persistSessionStateFromMethod: Re-read the record from persistence instead
of using a stale pre-await snapshot. Prevents race conditions where
concurrent session/update events (processed by persistSessionStateFromEvent)
are silently overwritten by optimistic updates.
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
* fix: correct doc examples with valid Codex modes and update stable API list
- Replace invalid Codex mode values ("plan", "build") with valid ones
("auto", "full-access") in agent-sessions.mdx and sdk-overview.mdx
- Update CLAUDE.md stable method enumerations to include new session
config methods (setSessionMode, setSessionModel, etc.)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: add OpenAPI annotations for process endpoints and fix config persistence race
Add summary/description to all process management endpoint specs and the
not_found error type. Fix hydrateSessionConfigOptions to re-read from
persistence after the network call, and sync mode-category configOptions
on session/update current_mode_update events.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Three independent fixes for the OpenCode agent adapter:
1. Wrong API endpoints: /event/subscribe → /event, /session/{id}/prompt → /session/{id}/message
2. Untagged enum mis-dispatch: replace serde_json::from_value with manual type-field dispatch
3. Wire permissionMode "bypass" for OpenCode: allow in normalize_permission_mode() and pass
--dangerously-skip-permissions to CLI (both spawn and spawn_streaming)
Tested with OpenCode 1.1.48 + Kimi K2.5.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Adds support for system, user, assistant, and result message types to the AMP schema, along with associated fields like subtype, session_id, tools, and duration metrics. Updates the schema validation and adds corresponding test cases. Also improves the command-line argument handling in the agent management package to accommodate the new message types and streamlined permission flags.
The changes enhance the schema's flexibility for different interaction patterns and provide better tracking of agent operations.
- Add Cursor to AgentId enum
- Implement install_cursor() function for binary installation
- Add Cursor spawn logic with JSON format support
- Update README to mention Cursor support in all relevant sections
Cursor-agent runs on localhost:32123 and uses OpenCode-compatible format.
Based on opencode-cursor-auth pattern for Cursor Pro integration.
Resolves#118
* chore: remove inspect.sandboxagent.dev in favor of /ui/
* chore: add 404 page
* fix: correct inspector package name in Dockerfiles and add .dockerignore
- Change @anthropic-ai/sdk-inspector to @sandbox-agent/inspector in all Dockerfiles
- Add .dockerignore to exclude target/, node_modules/, etc from Docker context
The wrong package name caused pnpm install --filter to match nothing, so the
inspector frontend was never built, resulting in binaries without the /ui/ endpoint.
* chore: cargo fmt
* chore(release): update version to 0.1.4-rc.7
The spawn_streaming() function was not passing environment variables
from SpawnOptions.env to the spawned process. This caused agents like
Claude to not receive ANTHROPIC_API_KEY, resulting in silent
authentication failures.
The non-streaming spawn() method correctly passes env vars (lines 298-300),
but spawn_streaming() was missing this code path.
This fix adds the same env var loop to spawn_streaming(), ensuring that
credentials extracted from the host environment are properly passed to
spawned agents.
* feat: add Claude adapter improvements for HITL support
- Add question and permission handling for Claude sessions
- Add Claude sender channel for interactive communication
- Add stream event and control request handling
- Update agent compatibility documentation
* fix: restore Claude HITL streaming input and permission handling
- Add streaming_input field to SpawnOptions for Claude stdin streaming
- Enable --input-format stream-json, --permission-prompt-tool stdio flags
- Pipe stdin for Claude (not just Codex) in spawn_streaming
- Update Claude capabilities: permissions, questions, tool_calls, tool_results, streaming_deltas
- Fix permission mode normalization to respect user's choice instead of forcing bypass
- Add acceptEdits permission mode support
- Add libc dependency for is_running_as_root check
Previously used cfg!(target_env = "musl") which checks compile-time,
causing musl-compiled sandbox-agent to always download musl binaries
even on glibc systems like Debian/E2B.
Now checks for /lib/ld-musl-*.so.1 at runtime to detect the actual
system libc and download the correct Claude binary variant.
- Add test_permission_mode() helper to use "default" for OpenCode
(it only supports default, not bypass or plan modes)
- Fix OAuth health check to accept 403 "Missing scopes" as valid auth
(OAuth tokens may lack api.model.read scope but still work for agents)
- Skip OpenCode in approval_flow_snapshots (doesn't support plan mode)
- Make sessions_list_global snapshot agent-agnostic (just check count)
- Add new snapshots for Codex and OpenCode agents
