OpenClaw's subconscious — the agent that learns from every session.
You correct your AI. It says "got it." Next session, same mistake.
memory-distiller fixes this.
Automatically scans every conversation for corrections, preferences, and insights. Applies a quality gate. Writes structured entries to persistent memory. Your agent gets smarter every time you use it.
npx clawhub@latest install memory-distillerZero config. Restart OpenClaw session after install.
| Signal | Trigger | Action |
|---|---|---|
| 🔴 Correction | "wrong", "no I meant", "actually..." | Records the correct rule |
| 💚 Preference | "I prefer", "always use", "from now on" | Saves your habit |
| 💡 Insight | "turns out", "the issue was", "solved" | Writes the lesson learned |
| 📌 Explicit | "remember this", "save this" | Immediately records |
Only records content that passes all 4 gates:
- Still valuable in 24 hours
- Reusable rule (not a one-off)
- Not already in
MEMORY.md - Can guide future behavior concretely
proactive-agent = Butler (captures present details, proactive behaviors)
memory-distiller = Historian (distills past conversations into lasting wisdom)
Install both for maximum effect:
npx clawhub@latest install proactive-agent
npx clawhub@latest install memory-distillerInspired by Claudeception, rebuilt for OpenClaw's architecture.
An agent should never be tripped up by the same problem twice.
Designed with 🐝 claw-multi-agent