HART OS is an open-source, AI-native operating system designed to run advanced AI models locally on any device, eliminating the need for datacenters.

It addresses the "capital lock" and "silicon lock" by enabling local inference through various hardware accelerators and supporting diverse hardware configurations. HART OS allows devices to recruit peers directly for computation, without intermediaries. The system is designed to adapt to the machine's capabilities, serving AI models like LLMs, vision, and speech to applications via a "Model Bus." It operates on a single Python codebase, adaptable to laptops, servers, and edge nodes, and federates peers directly over Web Sockets.

Key features include self-improvement capabilities, with user-controlled toggles for self-evolution and killable nodes. The project is in public alpha, seeking contributions for areas like the auto-evolve loop and guardrails. Installation requires Python 3.10 or 3.11 and involves cloning the repository and installing dependencies, with the core runtime listening on :6777.