Jan

Jan is an open-source desktop app that lets you download and run large language models like Llama, Qwen, and DeepSeek entirely on your own machine so your prompts and conversations never leave your device. Because inference happens locally, there is no account requirement and models keep working with the network turned off, which sidesteps the data collection baked into most hosted chatbots. It ships a built-in OpenAI-compatible server at localhost:1337, letting other apps use your local models without routing anything through a third party. Note that Jan can also connect to cloud providers such as OpenAI and Anthropic if you add API keys and doing so shares that data with those services so keep to local models for full privacy. It is not end-to-end encrypted in the messaging sense; the privacy benefit comes from on-device processing rather than encrypted transport. Built by Menlo Research and released under the Apache-2.0 license, it is actively developed and easy to audit.