Memory Overview (OpenClaw Docs)
Memory Overview (OpenClaw Docs)
Source: https://docs.openclaw.ai/concepts/memory
Storage layout
Memory is plain Markdown — no hidden state:
| File | Purpose |
|---|---|
MEMORY.md |
Long-term facts and preferences (always loaded) |
memory/YYYY-MM-DD.md |
Daily notes; today's and yesterday's loaded automatically |
DREAMS.md |
Optional dream diary and consolidation summaries |
Core agent tools
memory_search— semantic search over notes; finds relevant content even when phrasing differs.memory_get— reads a specific file or line range directly.
Search backends
| Backend | Description |
|---|---|
| Builtin (default) | SQLite-based; no extra dependencies |
| QMD | Local-first sidecar; adds reranking and query expansion |
| Honcho | AI-native cross-session memory with user modeling |
When an embedding provider is configured, search combines vector similarity with keyword matching.
Additional features
Memory Wiki Plugin
Compiles durable knowledge into a wiki vault with structured claims, contradiction tracking, and generated dashboards.
Automatic Memory Flush
Before compaction, the system reminds agents to save important context to files — ensures nothing critical is lost in a context window collapse.
Dreaming
Optional background consolidation process. Promotes qualified short-term signals to MEMORY.md based on a scoring threshold and recall frequency. Runs asynchronously.
Grounded Backfill
Replays historical notes to inspect durable candidates without directly promoting them to MEMORY.md. Useful for auditing what would have been retained.