#742 · Primary category: AI Agents & Automation

MetaClaw

agent ai-agent continual-learning fine-tuning llm lora meta-learning metaclaw online-learning openclaw reinforcement-learning skill-learning tinker

🦞 Just talk to your agent — it learns and EVOLVES 🧬.

Project last updated:06/07/26

GitHub Stars

3.5K

Forks

454

Contributors

21

License

MIT

Why we included this project

MetaClaw is a runtime for personal agents that gets better the more you use it. You run it in front of an existing agent like OpenClaw, and it pulls reusable skills out of each conversation, remembers facts and preferences between sessions, and schedules reinforcement-learning weight updates for idle windows, so the agent improves without interrupting active use. For anyone running a personal or team assistant, that solves the problem of static prompts that never reflect how you actually work. Setup stays out of the way: two CLI commands get a skills-only mode running against any OpenAI-compatible API, no GPU cluster needed, and the RL and scheduler features are optional extras. The async design keeps things responsive while scoring and training run in the background.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category