qhkm/zeptoclaw

Fast, small, secure, local-first personal AI assistant infrastructure: one Rust binary for tools, memory, channels, providers, and sandboxed autonomy.

What it solves

ZeptoClaw provides a lightweight, secure, and local-first infrastructure for running personal AI agents. It addresses the high resource overhead and security risks associated with AI agents that execute code, offering a high-performance alternative to larger frameworks that often lack built-in, multi-layered security guardrails.

How it works

Built as a single Rust binary, ZeptoClaw integrates LLM providers, tool execution, and memory management into a compact footprint. It uses a six-layer security model—including multiple sandbox runtimes (like Docker and Landlock), prompt injection detection, and secret leak scanners—to ensure that agent autonomy is safely contained. It can be deployed as a CLI tool, a multi-channel gateway (supporting Telegram, Slack, and Discord), or embedded as a library for other applications.

Who it’s for

It is designed for developers and power users who want to run AI agents locally, at the edge, or on a VPS, especially those requiring strict security isolation and minimal system resource usage (running on hardware as cheap as $10).

Highlights

  • Ultra-lightweight: A ~6MB binary with ~50ms startup time and ~6MB RAM usage.
  • Extensive Provider Support: Compatible with 18 LLM providers, including OpenAI, Anthropic, and local options like Ollama.
  • ** uma 33 Built-in Tools:** Includes capabilities for shell access, filesystem operations, web fetching, and more.
  • Robust Security: Built-in SSRF prevention, shell blocklists, and a tool approval gate to prevent dangerous executions.
  • Multi-Channel Integration: Acts as a gateway for various chat platforms including WhatsApp, Slack, and Discord.

Related

  • Project
    zeroclaw-labs/zeroclaw

    A self-hosted AI agent runtime written in Rust that connects various LLM providers to multiple communication channels and local system tools while prioritizing user data ownership.

  • Project
    microclaw/microclaw

    A self-hosted Rust agent platform providing a durable execution runtime for long-running AI tasks across a server-based automation backend and a native desktop workspace.

  • Project
    fastclaw-ai/fastclaw

    A lightweight AI agent runtime written in Go that allows users to create, manage, and deploy multi-agent systems with isolated sandboxes and multi-channel support.

  • Project
    librefang/librefang

    An open-source Agent Operating System built in Rust that enables the deployment of autonomous, scheduled AI agents that work independently of manual prompts.

  • Project
    nullclaw/nullclaw

    A tiny, high-performance AI assistant infrastructure written in Zig that enables autonomous agents to run on minimal hardware with near-zero overhead.