Jan

Jan is an open-source desktop application for chatting with large language models on your own computer or through a configured cloud provider. The Homebrew cask installs Jan.app from the official universal macOS ZIP, supporting both Apple-silicon and Intel Macs.

Local and connected AI

Download models from Hugging Face and run them locally, including supported Llama, Gemma, Qwen and GPT-oss models. Local inference lets you keep conversations on your machine when you choose local-only features. Cloud integrations include OpenAI, Anthropic, Mistral, Groq and other providers; requests to these providers leave your machine and require their credentials and service terms.

Assistants and integrations

Create specialized assistants for different tasks. Model Context Protocol integration connects tools for agentic workflows. An OpenAI-compatible local server, documented at localhost:1337, lets other applications use Jan's models. Version 0.8.4 adds native web search and fetch and a translating API gateway. Connected tools and web features are not offline-only operations.

Requirements and setup

The official README recommends macOS 13.6 or newer, with 8 GB RAM for 3B models, 16 GB for 7B and 32 GB for 13B models. Actual resource use varies with model and context size. Install the app, select and download a local model, or configure a cloud provider. The Node.js, Yarn and Rust requirements in the repository are for building from source, not for running the packaged application.

Version 0.8.4 migration

Settings move from webview localStorage into a backend-managed store, with secrets in the OS keyring. Existing settings and API keys migrate automatically at first launch. Old localStorage remains a downgrade snapshot; changes made in 0.8.4 or later do not update that old snapshot. Jan is licensed under Apache-2.0.

Sources

• Official repository and README
• macOS installation documentation
• Official 0.8.4 release
• Homebrew cask