Spotify Voice MCP

I built a tiny personal MCP for Spotify today because I wanted a more direct way to talk to my music. Not a shiny new interface to babysit—just the ability to say “play this,” “what’s on,” or “turn it down” and have it happen.

It uses Spotify’s Web API and stays local on my machine. The controls are deliberately small: search and play a song or artist, queue something up, pause, skip, check what’s playing, choose a device, and set the volume.

The interesting part was the connective tissue: a local OAuth loop, a one-time approval, and a refresh credential stored in Keychain instead of inside the project. After that, the voice layer could see my speakers and send the right command to whichever one was active.

I like that it feels less like building an app and more like making a tiny spell for my own room: “Play some BLACKPINK.” “Pause the music.” “Set the Boom box to fourteen percent.”