Skip to main content

​
Advanced escape hatches

Not part of the recommended setup. config.thinking overrides the sub-agent’s thinking level (default "off", since active memory runs in the reply path and extra thinking time directly adds user-visible latency):
config.fastMode overrides fast mode only for the blocking memory sub-agent. Use true, false, or "auto"; leave it unset to inherit the normal agent, session, and model defaults. "auto" uses the recall model’s configured fastAutoOnSeconds cutoff:
config.promptAppend adds operator instructions after the default prompt and before the conversation context — pair it with a custom toolsAllow when a non-core memory plugin needs specific tool order or query shaping:
config.promptOverride replaces the default prompt entirely (conversation context is still appended afterward). Not recommended unless deliberately testing a different recall contract — the default prompt is tuned to return either NONE or compact user-fact context for the main model:

​
Transcript persistence

Blocking sub-agent runs keep their runtime transcript in the agent’s SQLite store. By default, OpenClaw removes the temporary sub-agent session rows after the run finishes and does not create a JSONL file. If cleanup crosses the recall deadline, a completed summary grounded in memory results can still be recovered as timeout_partial after cleanup settles. This works with temporary transcripts; persistTranscripts only controls debugging exports. Failed runs, failed cleanup, and unavailable memory results remain ineligible for timeout recovery. Recovered summaries are not cached or stored in session debug lines. To export those transcripts as JSONL artifacts for debugging:
Exported transcript artifacts go under the OpenClaw state directory, in a plugin-owned, per-agent directory separate from active runtime state:
Agent ids are URI-encoded in this path: for example, support/agent becomes support%2Fagent. Change the final artifact subdirectory with config.transcriptDir. Use this carefully: exports can accumulate quickly on busy sessions, full query mode duplicates a lot of conversation context, and these artifacts contain hidden prompt context plus recalled memories.