Compaction

Every model has a context window (the maximum tokens it can see). Long-running chats accumulate messages and tool results; once the window gets tight, OpenClaw compacts older history to keep it within limits.

Compaction summarizes older conversation into a compact summary entry while keeping recent messages intact. The summary is stored in the session history, so future requests use:

- The compacted summary

- The latest messages after the compaction point'
''
'Compaction ''persists'' in the session's JSONL history.

See ''compaction configuration and modes'' for ''agents.defaults.compaction'' settings.

When a session approaches or exceeds the model's context window, OpenClaw triggers auto-compaction and may retry the original request with compacted context.

You'll see:

- ''🧹 Auto-compaction complete'' in verbose mode'
'- ''/status'' shows ''🧹 Compactions: <count>''

Before compaction, OpenClaw can run a ''silent memory flush'' to store persistent records to disk. See ''Memory'' for details and configuration.

Use ''/compact'' (optionally with a directive) to force a compaction pass:

/compact Focus on decisions and open questions

Context windows are model-specific. OpenClaw uses model definitions from the configured provider directory to determine limits.

- Compaction: Summarizes and persists in JSONL.

- Session pruning: Trims old tool results only, on-demand, in-memory.

See ''/concepts/session-pruning'' for details on pruning.

- Use ''/compact'' when a session feels stale or context-bloated.

- Large tool outputs are truncated; pruning can further reduce tool result buildup.

- If you need a fresh session, ''/new'' or ''/reset'' starts a new session ID.