Agents

You can run autonomous coding agents against a model served locally by Jan — keeping inference on your own hardware while using the agent tooling you already know.

The in-app OpenClaw integration was removed in Jan v0.7.9 due to resource usage and security concerns. Jan no longer runs agents inside the desktop app. Instead, agents run as external programs that you point at a local Jan model via the CLI, as described below.

How It Works

The jan launch command starts a local model server and then launches an agent CLI with everything pre-wired:

Jan starts the model and serves it on a local OpenAI- and Anthropic-compatible endpoint.
Jan sets the agent's connection details automatically — environment variables like OPENAI_BASE_URL / ANTHROPIC_BASE_URL, the API key, and the model name (for OpenClaw, Jan writes a provider entry to ~/.openclaw/openclaw.json instead).
The agent runs in your terminal, sending its requests to your local model.

The agent itself (Claude Code, OpenClaw) is a separate tool you install yourself — Jan only supplies the model backend and the wiring.

Launch an Agent


jan launch claude                                       # pick a model, then run Claude Code against it
jan launch claude --model janhq/Jan-code-4b-gguf        # use a specific model with Claude Code
jan launch openclaw --model janhq/Jan-code-4b-gguf      # wire OpenClaw to a local model

Run jan launch with no agent to pick one interactively. See the Jan CLI documentation for full options (port, context size, GPU layers, and more).

⚠️

Install the agent CLI first. For example, install Claude Code (opens in a new tab) or OpenClaw per their own instructions — jan launch runs them, but doesn't install them. (If OpenClaw isn't installed, Jan falls back to running it via npx.)

MCP Servers Projects