zeroclaw

ZeroClaw is a lightweight, security-first autonomous AI agent framework built in Rust. It is designed for high-efficiency performance with fast startup times, low memory consumption, and production-grade extensibility. Using a trait-driven architecture, it allows infrastructure teams to swap model providers, messaging channels, and memory systems without vendor lock-in. ZeroClaw emphasizes security with built-in sandbox controls, filesystem scoping, and encrypted secrets. It supports over 22 provider integrations, including OpenAI-compatible endpoints and local model workflows like Ollama, making it suitable for both server-class hardware and low-power devices like Raspberry Pi.

## How to Use

To use ZeroClaw, first install the Rust toolchain. Clone the repository from GitHub, build the release version using 'cargo build --release', and install it to your system path. Run the interactive onboarding via 'zeroclaw onboard --interactive' to configure providers and security settings. Finally, launch the agent as a background process using 'zeroclaw daemon'.

## Key Features

- 100% Rust-native architecture for high performance - Minimal runtime footprint (~3.4 MB binary, ~7.8 MB peak RSS) - Fast startup (<0.38s cold, <10ms warm) - Security-first defaults with sandboxing and allowlist controls - 22+ provider integrations including local model support - Hybrid memory system using SQLite (Vector + FTS5 keyword retrieval) - Multi-channel support (CLI, Telegram, Discord, Slack) - AIEOS identity profile support for consistent agent personas

## Use Cases

- Running 24/7 autonomous agents on low-power hardware like Raspberry Pi - Automating cross-platform bot workflows for Telegram, Discord, and Slack - Local developer automation for code and system operational tasks - Long-running automation on constrained server infrastructure