janhq/jan

Jan is an open-source desktop application that lets you download and run large language models, the kind of AI that powers ChatGPT, entirely on your own computer, with no internet connection required once the models are downloaded. The problem it solves is privacy and control: services like ChatGPT send your conversations to external servers, whereas Jan keeps everything local so your data never leaves your machine. With Jan, you browse and download models from HuggingFace, a platform that hosts open-source AI models. Once downloaded, models run locally using your computer's CPU or GPU. You can then chat with them through a user-friendly interface similar to ChatGPT. Jan also supports connecting to cloud AI providers like OpenAI, Anthropic Claude, Mistral, and Groq if you prefer to use remote models for tasks requiring more power than your local hardware can provide. Beyond basic chat, Jan runs a local OpenAI-compatible API server on your machine at localhost:1337, which means other applications can send requests to it as if it were the real OpenAI API. This enables developers to build or test AI-powered apps against locally running models before using the paid cloud API. Jan also supports the Model Context Protocol (MCP), which allows the AI to take actions, browsing the web, reading files, or calling tools, rather than only generating text. Someone would use Jan when they want a private, offline AI assistant, when they are developing AI applications and want to cut costs during testing, or when they simply want to experiment with open-source models without a subscription. It runs on Windows, macOS, and Linux. The tech stack is TypeScript for the application frontend, Tauri (a Rust-based framework) for the native desktop shell, and llama.cpp under the hood for running the actual AI models.