Avatar is an AI harness for running your agents in controlled environemnts. Avatars are agents that primarily interact with real users and the physical world.

Avatar Documentation

Project stage: Alpha.

Use local ai in your desktop or in your private cloud, with no worries about privacy or data leakage.

Chat with an ai agent locally. Build Agents with white list controls. Run Agents in safe environments.

Avatar has a very small footprint, it takes very low cpu resources. The docker image is also small.

Avatar can be run using

• Docker
• From Prebuilt Binaries
• Using Go Install
• Build and Run from Source

Avatar provides a user friendly and clean UI, accessible via your favourite web browser.

Avatar does not come with LLM or Image Processing models for now. Use external providers on your local machine like Ollama, MLX Serve, Docker Model Runner, llama.cpp

On mac, Avatar works best with MLX Serve.