# Off Grid AI

A local-first AI runtime and studio for macOS and Windows that runs open text, vision, image, and voice models entirely on-device behind one OpenAI-compatible API.

Off Grid AI is a local-first AI runtime for the desktop that downloads open models from a built-in catalog, or any GGUF from Hugging Face, and runs them across text, vision, image, and voice through bundled llama.cpp, stable-diffusion.cpp, whisper.cpp, and Kokoro engines. Nothing routes through a server the developer operates, so conversations, files, and models stay on the user's device.

It combines three parts: a chat studio with text and vision chat, on-device image generation, voice input and output, live artifacts, and retrieval-augmented projects; a gateway that exposes one local OpenAI-compatible API for chat, vision, image, and audio, which can also run headless; and an opt-in Pro layer that reads a user's screen through OCR, remembers it, and can act with approval. MCP connectors let it use external tools from within chat.

It runs on macOS and Windows, needs no account or cloud connection, and a companion mobile app offers the same on-device functionality on a phone.

Website: https://github.com/off-grid-ai/OGAD