## for OpenClaw
Long-term memory for OpenClaw
Composited search, auto-capture, and ten bundled memory skills the agent uses out of the box.
Long-term memory is the gap
OpenClaw agents are powerful in the session. The hard part is keeping what they learn around afterward.
- —Agents that only have raw logs can't actually recall. They need memory they can write to and query by meaning.
- —Bolt-on memory is opaque. You can't read or correct what the agent knows.
- —Knowledge from one OpenClaw run is trapped in that run.
- —Switching agents or vendors means starting over from scratch.
Basic Memory changes that
Give OpenClaw a persistent, searchable knowledge base it can read and write to.
Composited memory search
memory_search queries three sources in parallel: MEMORY.md (text + context), the full knowledge graph (hybrid full-text + vector), and active tasks in memory/tasks/. The agent gets results in clear sections so it knows where each piece of context came from.
Auto-recall, auto-capture
On every session start the plugin loads active tasks and recently modified notes. After every turn it captures the exchange as a timestamped entry in a daily note. Long-lived MCP connection means no cold starts per tool call.
Ten bundled memory skills
memory-defrag, memory-ingest, memory-lifecycle, memory-literary-analysis, memory-metadata-search, memory-notes, memory-reflect, memory-research, memory-schema, memory-tasks. The agent doesn't just have tools, it has named workflows for the recurring memory operations.
Plain Markdown you own
Everything lives as files in your project directory. Read them, edit them, version them. The same knowledge graph is available to Claude, ChatGPT, Cursor, and any MCP client — no vendor lock-in.
One command to get started
Quick Setup
openclaw plugins install @basicmemory/openclaw-basic-memoryThen enable the slot: `openclaw plugins enable openclaw-basic-memory --slot memory` and restart the gateway. The plugin auto-installs the bm CLI on first startup if it isn't already on PATH. Zero-config defaults work out of the box; see the integration README for full options. Verify with `openclaw plugins doctor`. Full setup guide →
Frequently asked questions
- How do I give OpenClaw agents persistent memory?
- Install the OpenClaw plugin with one command:
openclaw plugins install @basicmemory/openclaw-basic-memory, then enable the slot withopenclaw plugins enable openclaw-basic-memory --slot memoryand restart the gateway. The plugin auto-installs the bm CLI on first startup if needed. The agent gets composited search across MEMORY.md, the knowledge graph, and active tasks; plus auto-recall on session start, auto-capture per turn, and ten bundled memory skills. - What does the OpenClaw plugin actually add?
- A persistent MCP connection (no cold starts per tool call) plus fourteen agent tools (memory_search, memory_get, search_notes, read_note, write_note, edit_note, delete_note, move_note, build_context, list_memory_projects, list_workspaces, schema_validate, schema_infer, schema_diff), seven slash commands (/remember, /recall, /tasks, /reflect, /defrag, /schema, /bm-setup), and ten bundled memory skills the agent uses out of the box.
- How does the composited search work?
- When the agent calls memory_search, three sources are queried in parallel: MEMORY.md (text search with surrounding context), the full knowledge graph (hybrid full-text + vector search across all indexed notes), and active tasks (scans memory/tasks/ for in-progress work). Results come back in clear sections so the agent knows where each piece of context came from.
- Is the OpenClaw plugin open source?
- Yes, MIT-licensed and shipped as
@basicmemory/openclaw-basic-memoryon npm. Source lives in the public basic-memory repo underintegrations/openclaw/. Basic Memory itself is AGPL-3.0 and free to run locally; Basic Memory Cloud adds synced, hosted access from $15/seat/month with a 7-day trial.
