Spotify MCP

Spotify MCP is a lightweight Python server, built on the well-known spotipy library, that exposes the Spotify Web API to any Model Context Protocol client — Claude Desktop, Cursor, Zed, or the MCP Inspector. Once wired into your client's config with a Spotify developer Client ID/Secret and a loopback redirect URI, the model gains a first-party set of tools for controlling the currently active Spotify device and manipulating the user's library.

The tool surface is deliberately narrow and practical: start/pause/skip playback, search across tracks, albums, artists and playlists, fetch metadata for any of those objects, view and modify the play queue, and create or update playlists (add tracks, rename, reorder). That gives an LLM everything it needs to run natural-language DJ workflows — "queue three more songs like this one", "build me a 90-minute focus playlist from my liked songs", "skip anything by X for the rest of the session" — without you leaving the chat window.

Typical users are hobbyists and MCP tinkerers who want a concrete, useful demo of what agent tool-use looks like on their own data, plus power listeners who prefer a conversational interface to Spotify's app. It runs locally via uvx or a cloned checkout and speaks stdio to the client — nothing is proxied through a third-party server, and your Spotify tokens stay on your machine. The project is MIT-licensed with 600+ GitHub stars and contributions from several community maintainers.

A March 2026 notice on the repo flags the project as inactive: most pull requests will not be merged, and several recommendation-oriented endpoints the author had hoped to wrap have been deprecated by Spotify itself. What's shipped works, but do not expect new features.