Mux MCP

Mux's own MCP server brings Mux Video, Mux Data and Mux Robots into your agent. Instead of hundreds of endpoint tools, it gives your agent two: search_docs, to look things up in Mux's API and SDK docs, and execute, where it writes TypeScript against the full Mux SDK (@mux/ts) and gets back whatever the code returns. What you can do with it - Upload videos and manage assets, playback IDs, subtitle tracks and live streams - Get playback URLs and check the status of your assets - Run Mux Robots jobs, such as generating chapters, or summarizing a video and suggesting tags - Query Mux Data: views, performance metrics, errors, dimensions and filters - Look up answers in Mux's API and SDK documentation - Verify webhook payloads and manage signing keys How it connects Mux hosts it at https://mcp.mux.com. Agents that support OAuth sign you in to the Mux Dashboard and ask which environment to connect. Otherwise, create an access token in the Dashboard and send it as Basic credentials in the Authorization header. On the hosted server, each execute call runs in a fresh, isolated sandbox that can only reach Mux's API. To run it yourself, use npx -y @mux/mcp@latest with MUX_TOKEN_ID and MUX_TOKEN_SECRET set. Locally, the code runs in a Deno sandbox on your machine (Deno 2.9 or newer). Claude Desktop can also install it from the mux.mcpb bundle in Mux's releases. Read-only when you want it What it can do follows the environment and access token you authorize, and it doesn't treat destructive operations specially. To make sure your agent can't change anything, connect with an access token that has only Read permissions. When you run it locally, --code-allow-http-gets limits it to read requests, and other flags allow or block specific SDK methods or turn off either tool.