Skip to content

Introducing HomeClaw

HomeClaw is an AI assistant that runs on your machine. One installation is one agent: the same memory (RAG + agent memory), tools, and plugins no matter how you connect. Use cloud models (OpenAI, Gemini, etc.) or local models (llama.cpp, GGUF), or both together for better capability and cost.

Highlights

  • Companion app — Flutter-based app for Mac, Windows, iPhone, and Android (clients/HomeClawApp/): chat, voice, attachments, and Manage Core (edit core.yml and user.yml from the app). Makes HomeClaw much easier to use from any device.
  • Memory — RAG (vector + relational + optional graph) and agent memory: AGENT_MEMORY.md (long-term), daily memory (short-term). Cognee (default) or Chroma backend.
  • Plugins — Built-in (Python in plugins/) and external (any language: Node.js, Go, Java, Python, etc.). The system plugin (e.g. homeclaw-browser) is one external plugin written in Node.js; you can write plugins in any language and register them with Core. There are many tools and ecosystems you can leverage.
  • Skills — Full support for OpenClaw-style skillset: workflows in skills/ (SKILL.md); LLM uses tools and optional run_skill to accomplish tasks.
  • Multi-agent — Run multiple HomeClaw instances (different ports/configs); each is one agent. No special orchestration—just run more instances.
  • Cloud & multimodal — Gemini and other cloud models work well. Multimodal (images, audio, video) is supported with both local models (e.g. Qwen2-VL with mmproj) and cloud (e.g. Gemini, GPT-4o). Tested with both; all work well.

Core ideas

  • Cloud and local models — Use cloud models (LiteLLM: OpenAI, Gemini, DeepSeek, etc.) or local models (llama.cpp, GGUF), or both; they can work together for better capability and cost. Use local-only to keep data at home, or cloud for scale and features.
  • Channels — Reach your assistant from the Companion app, WebChat, Telegram, Discord, WeChat, WhatsApp, email, CLI, and more.
  • Memory — RAG-based, per-user; agent memory (AGENT_MEMORY.md, daily memory) for long- and short-term context.
  • Plugins & skills — Extend with built-in and external plugins (any language) and OpenClaw-style skills.

For the full intro and design details, see IntroducingHomeClaw.md in the repo.