Jan

Jan gives users a ChatGPT style chat interface for language models running entirely on their own machine. Built on Tauri with llama.cpp as its local engine, the open source desktop app downloads GGUF models such as Llama, Gemma, and Qwen from Hugging Face and runs them offline, with GPU acceleration on NVIDIA, AMD, and Intel Arc hardware. When cloud access is wanted, the same interface connects to OpenAI, Anthropic, Mistral, Groq, and other providers. Jan exposes an OpenAI compatible API server on localhost:1337 so other applications can use the local models, supports custom assistants, and integrates the Model Context Protocol for agentic tool use. Builds are available for Windows, macOS, and Linux. Privacy conscious users and developers who want local inference without giving up a polished chat interface are the primary audience. The application is released under the Apache 2.0 license.