diff --git a/packages/coding-agent/docs/compaction.md b/packages/coding-agent/docs/compaction.md
index a1753a4e..168ce198 100644
--- a/packages/coding-agent/docs/compaction.md
+++ b/packages/coding-agent/docs/compaction.md
@@ -23,13 +23,13 @@ Auto-compaction triggers when:
 contextTokens > contextWindow - reserveTokens
 ```
 
-By default, `reserveTokens` is 16384 tokens. This leaves room for the LLM's response.
+By default, `reserveTokens` is 16384 tokens (configurable in `~/.pi/agent/settings.json`). This leaves room for the LLM's response.
 
 You can also trigger manually with `/compact [instructions]`, where optional instructions focus the summary.
 
 ### How It Works
 
-1. **Find cut point**: Walk backwards from newest message, accumulating token estimates until `keepRecentTokens` (default 20k) is reached
+1. **Find cut point**: Walk backwards from newest message, accumulating token estimates until `keepRecentTokens` (default 20k, configurable in `~/.pi/agent/settings.json`) is reached
 2. **Extract messages**: Collect messages from previous compaction (or start) up to cut point
 3. **Generate summary**: Call LLM to summarize with structured format
 4. **Append entry**: Save `CompactionEntry` with summary and `firstKeptEntryId`