Evidence record / claude-code

A dated record of one change, kept so the writing that cites it can be checked. Compiled from the sources listed below by the research run, not written for reading. The judgment lives in the issues that cite it, below.

2026-10-24-claude-code-idle-compaction-reduces-recaching-work-by-replacing-long-idle-context-with-a-summary

Idle compaction reduces recaching work by replacing long idle context with a summary. Eligible interactive terminal conversations on a one-hour prompt-cache TTL, with Anthropic feature-flag fetching available, compact after about 54 idle minutes once context reaches the configured threshold, 200,000 tokens by default, while the computer is awake. A normal 200K context window compacts before reaching that default idle threshold; lowering it can make the session eligible, with a 100,000-token minimum. idleCompaction: false disables it; CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS changes the threshold. The later settings UI makes the choice discoverable. On return, the model receives a summary instead of all previous context.

Channel: tagged-release. Half: both. Date: 2026-10-14.

Operator consequence: Native compaction can replace a timer or manual cleanup habit around long sessions. For a debugging thread where exact earlier details matter, choose the threshold or disable the behavior and retain external evidence. The expected saving is less full-context recaching; no measured quota reduction or accuracy result was established here. Do not use this mechanism as an explanation for the separate usage-drain complaints.

Receipt

Finding metadata

Run: 2026-10-24-weekly-digest-2026-10-13_2026-10-24-frontier-v0

Finding ID: 2026-10-24-claude-code-idle-compaction-reduces-recaching-work-by-replacing-long-idle-context-with-a-summary

Profile citations

Source links

Primary links, including exact changelog lines when available.

Versioned source: run artifact