zeroclaw-labs/zeroclaw

Project overview

The project addresses owning and running a fully autonomous AI personal assistant across any OS and platform without relying on third-party hosted infrastructure, packaged as a single fast and small Rust binary that can be deployed anywhere.

Deployment

Refer to project documentation

License

Apache-2.0

Best for

• Developers and AI engineers who want to run a provider-agnostic, multi-channel AI agent entirely on their own machine with their own keys.
• Operations teams that need event-triggered automation with approval gates, resumable runs, and always-on background service installation.

Key capabilities

• Inbound messages from Discord, Telegram, Matrix, email, webhooks, and CLI are delivered to a single agent loop.
• Interact with GPIO, I2C, SPI, and USB on Raspberry Pi, STM32, Arduino, and ESP32 via the Peripheral trait.
• HTTP/WebSocket gateway with a web dashboard for chat, memory browsing, config editing, and tool inspection.
• Event-triggered SOPs via MQTT, webhook, cron, or peripheral with approval gates and resumable runs.
• Register and run the agent as an always-on background service via systemd, launchctl, or Windows Service.

Limitations and risks

• Requires manual TOML configuration with at least four section headers for a minimal working setup.

Getting started

• Run the one-shot installer for Unix or use the prebuilt PowerShell path for Windows. Execute zeroclaw quickstart to automatically write a working config that configures a provider. Then run zeroclaw agent -a <alias> to start chatting. Coding is optional for setup.

Evidence and sources

Open original GitHub project