microclaw/microclaw

🦀An agentic AI assistant that lives in your chats, inspired by nanoclaw and incorporating some of its design ideas. Built with Rust 🦀

What it solves

MicroClaw is a self-hosted agent platform designed for complex, long-running tasks that extend beyond a single request. It addresses the need for durable execution, where agent sessions, tool use, and scheduled tasks can survive process restarts and network interruptions without losing progress.

How it works

Built in Rust with an embedded SQLite database, the project uses a shared Agent Engine that powers two different interfaces:

  1. MicroClaw Server: An always-on backend for chat channels (like Discord, Slack, and Telegram), APIs, and automation.
  2. MicroClaw Work: A native desktop application for local, project-centered agent work with Git and file system integration.

Every interaction follows a loop of resuming session state, calling a provider-neutral model, executing tools through guardrails, and durably persisting the result.

Who it’s for

It is intended for users who need a reliable, self-hosted AI agent runtime that can handle multi-step tool use, persistent memory, and scheduled automation across various platforms and communication channels.

Highlights

  • Durable Execution: Checkpoints and recovery models ensure work resumes from the last successful step after a crash or interruption.
  • Provider Agnostic: Supports Anthropic, OpenAI-compatible providers, Ollama, and other local LLMs.
  • Local-First Workspace: Native desktop integration for Apple Silicon macOS with workspace-aware file attachments and Git context.
  • Extensible Architecture: Supports MCP servers, plugins, hooks, and custom skills without modifying the core runtime.
  • Integrated Safety: Includes tool risk gates, Docker sandboxing, and egress controls built directly into the execution flow.

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
    qhkm/zeptoclaw

    A fast, secure, and local-first personal AI assistant infrastructure written in Rust, providing sandboxed autonomy and multi-provider LLM support in a single 6MB binary.

  • Project
    swarmclawai/swarmclaw

    A self-hosted AI agent runtime and multi-agent framework for creating autonomous agent swarms with durable memory, scheduling, and multi-provider LLM support.

  • 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
    Superagentsys/novalclaw

    A local-first AI agent platform with a Rust core and desktop interface that allows users to build personalized agents with MBTI-driven personalities and multi-channel deployment.