ZeroClaw is an agent runtime that deploys as a single Rust binary to run autonomous AI assistants on your own machine.
The tool addresses the need for AI agents that operate under user control rather than through third-party services. It works by running locally on your hardware, connecting to LLM providers such as Anthropic, OpenAI, and Ollama, and exposing the agent through multiple communication channels including Discord, Telegram, Matrix, email, voice, and webhooks. The agent can act through tools like shell commands, browser automation, HTTP requests, hardware control, and custom MCP servers. All execution happens in your workspace with your API keys, keeping data and computation local.
ZeroClaw suits developers and users who want full ownership of their AI assistant infrastructure without relying on cloud-hosted agent platforms. It works across operating systems including Linux, macOS, Windows, FreeBSD, and NixOS, with Docker support available. The project emphasizes that you own the agent, the data, and the machine it runs on. Those evaluating adoption should note that the tool is built in Rust and distributed as a prebuilt binary for most platforms, with guided installers available for Unix and Windows to simplify setup.
The project maintains an active build pipeline with passing continuous integration checks. Documentation is comprehensive, including a philosophy section explaining the project's core opinions, architecture documentation, and platform-specific setup guides. The codebase includes translated documentation managed through git submodules, indicating sustained internationalization effort. Installation tooling is automated through cargo-based generation of platform-specific installers, and the project provides multiple installation paths to accommodate different user preferences and technical backgrounds.
