zeroclaw-labs/zeroclaw

ZeroClaw is a self-hosted AI personal assistant runtime written in Rust. The core problem it solves is that most AI assistants either run in someone else's cloud (where your data and conversations leave your control) or are locked to a single interface. ZeroClaw is a single binary you install on your own machine, laptop, server, Raspberry Pi, or a Docker container, that connects to whichever large language model providers you choose and lets you talk to it through any of more than 30 communication channels simultaneously. How it works: you configure a single TOML file listing your preferred AI providers (Anthropic, OpenAI, a local Ollama instance, or any OpenAI-compatible endpoint), the channels you want to use (Discord, Telegram, Matrix, email, webhooks, or a terminal), and the tools the agent may use (shell commands, web browsing, HTTP requests, or custom external servers). An agent loop processes incoming messages from all connected channels, calls the configured model, and executes approved actions. A security policy controls how autonomous the agent is, in the default supervised mode, medium-risk operations require your approval before running. An SOP (Standard Operating Procedure) engine lets you define automated workflows triggered by events like incoming webhooks, cron schedules, or hardware signals from GPIO pins. You would use ZeroClaw if you want a personal AI assistant that you fully control, that works across every messaging platform you use, can take actions on your computer, and keeps all data on your own hardware rather than third-party servers. It also supports hardware integrations for Raspberry Pi and microcontrollers, making it useful for home automation or embedded projects. The tech stack is Rust (2024 edition), distributed as a compiled binary with optional source builds. Configuration is TOML; memory is stored in SQLite.