Auto-recall
Injects relevant memories into every prompt.
The LLM sees them as natural context — no "according to my records" phrasing needed.
OpenClaw Plugin
@keyoku/openclaw is the bridge between the Keyoku Engine and OpenClaw. It gives your AI agent persistent memory, proactive awareness, and explicit memory tools — all through a single npm package.
What it does
The plugin hooks into OpenClaw's lifecycle to add four capabilities to your agent, with zero manual wiring.
Auto-capture
Extracts structured facts from every user-agent exchange.
Incremental by default — captures per-message, not per-session.
Heartbeat integration
Detects heartbeat prompts from OpenClaw and calls the engine for a 14-signal scan.
Injects proactive signals when the agent should speak up.
7 LLM tools
Registers memory_search, memory_get, memory_store, memory_forget, memory_stats, schedule_create, and schedule_list as callable tools.
Architecture
The plugin hooks into OpenClaw via two lifecycle hooks: before_prompt_build and agent_end. It manages the keyoku-engine binary as a sidecar process — spawning it on startup, health-checking on /api/v1/health, and sending SIGTERM on shutdown.
| Hook | Purpose |
|---|---|
before_prompt_build | Stashes user input, injects recalled memories, and handles heartbeat signal injection. |
agent_end | Pairs the stashed user message with the agent response and sends the exchange to the engine for capture. Also records heartbeat response fingerprints for deduplication. |
Package details
| Field | Value |
|---|---|
| Package | @keyoku/openclaw |
| Registry | npm |
| Kind | memory |
| Engine | Managed sidecar (Go binary) |
Next steps
Get started with the plugin by following the guides below.
- Installation — setup wizard and manual install
- Configuration — every config parameter explained
- Hooks & Tools — hook lifecycle and LLM tool reference