ZeroClaw

ZeroClaw is a Rust-based AI assistant framework focused on very low resource use, fast startup, and production-style infrastructure for channels, providers, tools, memory, and security policy. Its public site positions it against heavier assistant gateways: less RAM, a small binary, trait-driven extension points, and support for many AI providers and chat channels. It is a good AgentRiot listing because it shows the opposite design direction from monolithic assistant gateways.

Top features

• Rust implementation: memory safety and a small compiled binary are central parts of the project’s pitch.
• Low resource target: the site advertises sub-5MB RAM, sub-10ms startup, and roughly 3.4MB binary size in its benchmark snapshot.
• Pluggable architecture: providers, channels, tools, memory, runtime adapters, and security policy are expressed through traits.
• Broad provider and channel support: the project lists OpenAI-compatible providers, Anthropic, Gemini, Ollama, Telegram, Discord, Slack, Matrix, Signal, WhatsApp, email, and webhook paths.
• Security controls: pairing, sandboxing, allowlists, encrypted secrets, filesystem scoping, and tunnel-only public access are documented.

Use cases

• Run an assistant gateway on edge hardware, small servers, robots, or air-gapped devices where a TypeScript or Python stack is too heavy.
• Build messaging bots that need provider switching, local memory, tools, and channel routing.
• Use a Rust framework for assistant infrastructure where binary size, startup time, and explicit trait boundaries matter.
• Compare lightweight Claw-family runtimes against OpenClaw, PicoClaw, NanoClaw, and Hermes.
• Prototype ops, SRE, legal, education, IoT, or API assistant workflows with a gateway designed for small machines.