Open-source desktop AI workspace for local models, chat, provider connections, and a local API, with separate agent tooling.
Jan
Explore features, practical uses and pricing below.
Jan provides a desktop interface for running language models locally and working with them through conversations. Its broader ecosystem includes model connections, a local API server, and agent tooling. The local workflow is useful for users who want to choose their model and keep inference on their own machine, while cloud connections remain a separate option.
Jan suits technically curious users, developers, and people evaluating local AI for everyday work. It is particularly useful when a graphical interface makes model selection easier than managing a command-line server alone. Users should distinguish Jan Desktop from the separately documented Jan Agent and choose the component that fits the intended task.
Download a model that fits your machine and test a short document-summary request. Check speed, memory use, and the quality of the response before moving to a larger model. If a local application needs inference, enable the documented local API and test a simple request. Keep cloud providers disabled for a workflow intended to remain local, and verify where any connected tool sends its data.
Local inference depends on available memory, model size, architecture, and supported hardware. A smaller model may be convenient but less capable for complex reasoning or tool use. Local execution also does not make every connected workflow private: remote providers and external tools can process data elsewhere. Review the selected model's license and each connection rather than treating the desktop application's open-source status as a universal deployment guarantee.
Jan's desktop software is available through its official downloads and documentation. Running local models uses your own hardware and storage, while external provider connections may introduce separate usage charges. Check the current installation requirements, model support, and component documentation before building a workflow around a particular API or agent capability.
Supported local models can run on your machine without a cloud provider key.
The official documentation distinguishes the desktop workspace from separately distributed agent tooling.