openclaw-mem

Govern memory, episode, skill, and fact lifecycle maintenance through the unified curate scan, review, apply, verify, and rollback verbs.

Operate graph readiness, topology lookup, synthesis-aware routing, and symbolic evidence canvases. Use for project ownership, impact, dependency, or idea-to-project questions.

Govern embedding refresh, episodic evidence synchronization, metadata writeback, and engine dataset snapshots. Use before or after bounded sync, reindex, migration, or writeback operations.

Govern source-linked temporal facts in openclaw-mem. Use for current truth, timelines, stale detection, fact packs, and review-only fact extraction.

Operate experimental Dream Lite planning, governed refresh canaries, rollback, and Director rehearsals. Use only for explicitly reviewed derived card maintenance.

Operate the experimental GBrain lookup, restricted helper-job, and governed refresh-canary lanes. Use only when GBrain integration is explicitly enabled.

Operate the experimental read-only goal primitive and surface validation. Use for normalized goal status, compact goal packs, and receipt validation, not continuation or autonomous mutation.

Inspect experimental self-improvement consolidation surfaces. Use for read-only validation, goal readback, staged skill proposals, curator review, and system status.

Inspect and govern the experimental derived self-model sidecar. Use for continuity snapshots, attachment maps, drift, threats, public-safe wording, release receipts, or migration comparisons.

Trust-aware OpenClaw memory routing. Use when an agent must decide whether to recall, search docs or graph evidence, build a bounded pack, store a durable fact, or keep information session-local.