Jan Docs
DocsJan DesktopCLI

Jan CLI and Desktop

Install the CLI separately

Since Jan Desktop 0.8.5, the desktop app no longer bundles or installs the jan CLI, or adds it to your PATH. There is no Desktop Settings > General > Jan CLI installation step.

The current CLI is Jan Agent, a terminal agent installed separately from Desktop. It is a preview, distributed through the agent-nightly channel, not a stable Desktop feature. Follow the Jan Agent quickstart for installation and provider setup, and the CLI reference for supported flags and subcommands.

After installing, check which version runs:

jan --version
jan --help

Plain jan opens the interactive agent and auto-approves tool calls. Use jan --safe for approval prompts. Shell sandboxing is a separate option, --sandbox; neither control is enabled by default.

Upgrading from a bundled CLI

The Desktop 0.8.5 migration notes describe what happens to older installations:

  • macOS and Linux: a CLI installed by Desktop 0.8.4 in /usr/local/bin or ~/.local/bin is left in place but stays at 0.8.4. Install the current Jan Agent separately. If the installer warns that another jan comes first on your PATH, resolve that conflict before using it.
  • Windows: the Desktop update removes the old jan.exe from the app's resources\bin folder. Install Jan Agent separately into %USERPROFILE%\.local\bin and follow the quickstart's -AddToPath instructions. The old app resources\bin entry may remain in your user PATH; you can remove that obsolete entry.

Updating Desktop does not update a separately installed Jan Agent. The desktop executable is now Jan-Desktop; it is not the jan terminal agent.

Local models and legacy commands

Older versions of this page documented a local-inference CLI with jan serve, jan launch, jan models load, and jan models load-mlx. Those are not commands in the current Jan Agent CLI. The current model and thread commands live under jan cli models and jan cli threads; jan cli models list lists models exposed by configured providers, rather than loading a Desktop model. Use the Jan Agent CLI reference instead of legacy command examples.

Jan Agent does not include a local inference engine. To use your own hardware, download a model in Desktop and start its Local API Server, then configure Jan Agent to use that OpenAI-compatible endpoint. See Jan Agent provider setup for local models.