TrackRate user guide
MCP AI Integration
Connect an MCP-compatible AI client to TrackRate for library analysis, metadata, tags, playlists, and playback.
6 questions
What is TrackRate MCP, and what can an AI do with it?
Premium MCP exposes approved TrackRate library and playback capabilities to a compatible AI client through the Model Context Protocol. An agent can inspect the library, search, read supported metadata and lyrics, update supported fields, manage tags, create and arrange playlists, and control playback.
MCP does not make the model infallible. Treat it as an operator with tools: give a narrow goal, ask for a preview or explanation, and verify important changes in TrackRate.
How do I connect an MCP client to TrackRate?
- Open Settings → Remote API.
- Turn on the API server.
- Enable MCP.
- Open the MCP gear.
- Copy the complete endpoint exactly as shown.
- Add it to the compatible client and keep TrackRate running.
MCP is local-only by design. A port mismatch, omitted path segment, disabled API server, disabled MCP toggle, or firewall rule can prevent the connection.
Can AI fix inconsistent artist, album, title, year, genre, comment, and tag metadata?
Yes, within the metadata fields TrackRate stores and exposes. A safer cleanup starts by asking the agent to inspect raw values, propose a mapping, and change a small batch. For example: “List inconsistent spellings of The Beatles, propose one canonical value, and update no more than 20 tracks after showing the plan.”
TrackRate does not expose every extended audio field for editing. Composer, album artist, disc number, conductor, publisher, copyright, and similar fields may be readable from files but are not available for AI correction through TrackRate.
What are good, specific prompts for TrackRate MCP?
“Find tracks from 2020 onward with an artist but no album, group them by folder, and do not edit anything.”
Tag cleanup“Compare #rock, #rock-n-roll, and #classic-rock. Propose a merge plan and wait for approval.”
Playlist“Find 30 tracks rated 8–10, tagged #road, between three and eight minutes, excluding #live. Create ‘Road Trip Draft’ and arrange it with gentle energy waves and artist spacing.”
Specific scope, limits, and stop conditions produce better results than “fix my whole library.”
| Instead of | Ask for | Safety boundary |
|---|---|---|
| “Fix my metadata” | Find post-2020 tracks missing albums and group them by folder | Inspect only; make no edits |
| “Clean my tags” | Compare three named tag variants and propose a merge map | Wait for approval |
| “Make a road playlist” | 30 tracks, ratings 8–10, #road, 3–8 minutes, excluding #live | Create a new draft playlist |
How should I run a large AI cleanup safely?
Work in stages of one artist, album, folder, or tag family. Ask the agent to keep a scratch summary of completed mappings and skipped cases so a later batch can resume consistently. Review each stage in TrackRate before starting the next.
MCP writes happen immediately and there is no staging area. TrackRate creates one restore point per MCP session, not one undo entry per edit. Create your own backup before a large job and avoid asking a single prompt to rewrite the entire library.
Why did an MCP-created playlist have an unexpected order?
Order matters in the tool sequence. If the final order is important, the AI should call arrangePlaylist before selectPlaylist. Selecting an unarranged playlist through MCP can trigger an automatic arrangement with a random strategy.
State the order requirement explicitly in the prompt: name the candidate filters, requested count, strategy, and whether the agent should select the playlist only after arrangement completes.
No questions match that search.