14b. Driving Nebula from an AI assistant (MCP)

Nebula ships a small MCP (Model Context Protocol) server in mcp/. It lets an assistant that speaks MCP open your project, set the tempo, start and stop playback, browse the instrument and device catalogues, and render — through the app's own Control API, with the same rules.

• Install and run: pip install -r mcp/requirements.txt, then python mcp/server.py --port 8642 --token <the token from Settings → API>. mcp/README.md has a configuration to copy into an MCP client.
• Rendering is one tool call: the server waits for the render to finish and tells the assistant where the file was written.
• One client at a time: while the server is connected, another script cannot use the API. That is the API's rule, not the server's.