Connect your own AI to the studio (MCP)
evmux has an MCP (Model Context Protocol) server. It lets you drive your evmux studio from an AI client you already use - Claude, ChatGPT, Cursor, or any client that supports MCP connectors - instead of the assistant built into the studio.
Add evmux as a connector
- In your AI client, add a custom connector / MCP server.
- Use this URL:
https://console.evmux.com/api/mcp - The client opens the evmux login and consent screen. Sign in and approve. There is nothing to copy or paste.
If your client asks for a token instead of a login, open https://console.evmux.com/actions/mcp/token while logged in to evmux. It shows a personal access token and the header to use (Authorization: Bearer ...).
Before you use it
Your studio must be open in the browser. The AI’s commands run inside that studio tab, so if the tab is closed the AI will tell you no studio is open. If you have several studios open, the AI asks which one to use and remembers your choice.
What the AI can do
Unlike the assistant built into the studio, which only answers questions, a connected AI can act:
- Read your scenes, layers, assets, participants and apps.
- Add, move, resize and edit layers, add text, images, videos and participants, switch scenes.
- Search the free stock library and import items.
- Answer questions and highlight the right control in the studio with “Show me”.
Reading never needs confirmation. Anything that removes something or changes what is on air, the AI asks you before doing it. Generating AI images is not available through the connector; use the image generator inside the studio.
Cost
Your own AI plan pays for its tokens. Using the connector does not use your evmux AI allowance. evmux rate limits and the AI image quota still apply.
While an external AI is working, the AI button in the studio shows the processing indicator.