AI Grid
A tool that networks computers a user already owns into a single OpenAI-compatible endpoint, routing each inference or training request to whichever machine runs the right model, with no account or relay required.
AI Grid puts every computer a user owns behind one OpenAI-compatible endpoint and routes each request to whichever machine runs the appropriate model. It is local by default: setting up a grid means running an install script and a start command on a host machine, then joining other machines that each run a model through an engine such as llama.cpp, with model files pulled directly from repositories like Hugging Face.
If a user already runs Ollama, vLLM or LM Studio, the grid networks them rather than replacing them, and machines in different physical locations can join in a remote mode. The project also supports fine-tuning workflows, including GRPO and LoRA, across the networked machines, and offers a desktop app in addition to its command-line interface.