Atomic Chat

Atomic Chat is an open-source chat interface you can run entirely on your own machine. It gives you the option to load models locally from HuggingFace or connect out to external providers, so you can keep things private or use cloud services when you need them.

• Local models — run LLMs such as Llama, Gemma, Qwen and others from HuggingFace on your hardware.
• Cloud integration — connect to OpenAI, Anthropic, Mistral, Groq, MiniMax and similar services.
• Custom assistants — build specialised assistants tuned to particular tasks.
• OpenAI-compatible API — a local server (localhost:1337) for other applications to call.
• MCP support — Model Context Protocol integration for agent-like behaviours.
• Privacy-first — you can run everything locally when you prefer.

It’s suitable if you want control over where models run and how data is handled, especially for people who need offline capability or like mixing local and cloud models. The project is free and open-source, so you can review or adapt the code if that matters to you.