Automation

Monet can run a small local server so AI agents can control the app over the Model Context Protocol. An agent connected this way can list projects, generate images, run flows and read the library.

Set it up under Settings → MCP Server:

1. Tick Enable MCP server.
2. Leave Port at 3600 unless something else needs it.
3. Choose Save MCP Settings. The server starts — and stops — as soon as you save.

Point your MCP client at http://127.0.0.1:3600/mcp. The server listens on localhost only, so nothing beyond your own computer can reach it.

Generations and flow runs an agent starts appear in the same task bubbles as ones you started yourself while they are running, so you can watch their progress, or cancel them, from the app.

MCP tools can also create projects, prompts, flows and scenes; upload and read images; inspect tasks; add, update or remove scene layers; and delete scenes. Interface tools can open a particular view, scene, flow, image or task, select a layer, and set the scene zoom and pan.

The same port also serves the app’s REST API. The monet command-line client uses it for projects, prompts, generation, library images, flows, scenes and layers. Run monet help for its commands; use --url or MONET_URL if you changed the port.