Jan
Jan is an open-source desktop application for running large language models locally and offline, or connecting to cloud models like GPT, Claude, and Gemini through API keys, built for privacy-focused users and developers who want their chat history and data to stay on their own machine.
Use Cases
Run open-source LLMs locally and offline
Chat with cloud models like GPT, Claude, and Gemini via API key
Connect MCP servers to give models tools
Keep chat history and data private on-device
Serve a local OpenAI-compatible API endpoint
Upload images and documents for multimodal models
Pros
Fully open-source under AGPL with auditable code on GitHub
Runs models offline so chat data stays on-device
Single interface for both local models and cloud API providers
Connect any MCP server with per-tool permission gating
Local OpenAI-compatible API server for plugging into other apps
Cons
Less refined model browser than LM Studio
Local model speed limited by your own hardware
Cloud models require bringing your own API keys
Smaller curated model catalog than some rivals
AGPL license can complicate commercial redistribution
Platforms
Desktop
API
MCP
What sets Jan apart from LM Studio, the polished default in local AI, is that you can read and fork the whole thing, and your chats live on disk as plain files rather than inside someone's app. That openness costs some polish, with a thinner model catalog and inference capped by whatever hardware you own.
Must Try