Favicon of PersonalClaw

PersonalClaw

Self-hosted personal AI agent with memory, scheduled automation and configurable local or cloud models. Runs on Linux, macOS or Docker under MIT.

Screenshot of PersonalClaw website

PersonalClaw is a self-hosted AI agent for one person who wants chat, ongoing tasks and personal knowledge in one web dashboard. It runs on Linux and macOS, with Docker support that also covers Windows. It's open source under MIT and is currently pre-1.0, with breaking changes expected. A one-time download supplies a tiny CPU model for offline chat without tools. Autonomous tool work needs your own Ollama model or a provider app.

You can give an agent a goal and let it plan and work through repeated cycles, with controls to pause, redirect or stop it. Chat supports tool approvals, separate sessions, conversation branches and temporary memory modes. Long-term memory learns from corrections and recalls relevant context; an optional Markdown vault works with Obsidian.

The knowledge base accepts documents, web pages and media, then brings semantic search and extracted relationships into chat. Reusable skills capture procedures, while scheduled jobs and webhooks trigger automation. Background agents can handle work, and a channel inbox watches messages and drafts replies. MCP support works in both directions.

Model providers, local models and channel connections are removable apps, so you aren't tied to one vendor. The app platform requires permission before activating additions. Execution controls include tool approvals, shell restrictions and policies for outbound connections.

Conversations, memory and knowledge stay on your machine unless you connect a remote provider. PersonalClaw sends no analytics, crash reports or usage tracking. It does contact GitHub for automatic update checks by default, which you can disable. State exports exclude credentials.

Similar to PersonalClaw