Music MCP Server for Claude: Tag Tracks in Chat

October 2, 2026·by TrackTag team

A music MCP server connects an AI assistant to tools that can work with audio, rather than asking the assistant to infer details from a filename or a prompt. TrackTag’s MCP server lets Claude Desktop, Cursor and Claude Code analyze a selected track, find local audio files by name and check your credit balance from a conversation. The MCP process runs on your computer; TrackTag performs the requested analysis through its service. (tracktag.me)

That setup is useful when you are reviewing demos, preparing a small delivery or checking a track while working in an AI assistant. It is not a substitute for a catalog-wide batch workflow. Below is a reproducible setup, prompts to try, and guidance on when to switch to Studio, Zapier or the API.

MCP music analysis: what the server does

MCP, or Model Context Protocol, is a way for an AI assistant to call external tools. Think of the assistant as the conversation interface and the MCP server as the connection to TrackTag. You ask for an action, the assistant calls the relevant tool, and the result comes back into the conversation.

With TrackTag, you can ask an MCP-capable assistant to:

  • Analyze a local audio file or a track available at a URL.
  • Find an audio file on your computer by name.
  • Check your remaining credits.
  • Review recent analyses.

The assistant can then summarize or organize the returned analysis in the way you requested. For example, ask it to list the mood, vocal description and production notes in a compact review. The underlying audio analysis comes from TrackTag, not from the language model guessing what a track might sound like. (tracktag.me)

TrackTag Studio returns up to 35 fields per track, including BPM, key, genres and subgenres, moods, emotions, themes, occasions, instruments, vocals, song structure, production and mixing notes, and a written description. Core returns 9 fields for 1 credit; Ultra returns all 35 for 2 credits. Both levels use the same engine on the same audio and are equally accurate. Core changes how much information you receive, not the quality of the analysis. You can upgrade a Core result to Ultra later for the 1-credit difference.

BPM and key are measured from the audio signal, rather than guessed by a language model. If a track does not have a steady beat, TrackTag can return No BPM instead of making up a tempo. That matters when a library is sorting material by tempo or when a producer is checking a track whose pulse is ambiguous.

Set up TrackTag in Claude, Cursor or Claude Code

Start by creating an API key in TrackTag Studio. In Studio, open the integrations or API and MCP area, create a key and copy it somewhere secure. Treat it like a password. The key is shown once when created, and it authorizes the assistant to use your TrackTag balance for paid analyses. (tracktag.me)

For Claude Desktop or Cursor, add the following entry to the MCP configuration file used by your app. If the file already has an mcpServers object, add only the tracktag section inside it. Replace the example key with your own.

```json

{

"mcpServers": {

"tracktag": {

"command": "npx",

"args": ["-y", "tracktag-mcp"],

"env": {

"TRACKTAG_API_KEY": "tt_live_YOUR_KEY"

}

}

}

}

```

Save the file and fully restart the assistant so it loads the server. In Claude Desktop, quitting the app completely matters; closing only its window may not reload the configuration. Cursor uses its MCP configuration file, commonly ~/.cursor/mcp.json. You need Node.js available so npx can run the TrackTag package.

For Claude Code, register the server from a terminal with:

```bash

claude mcp add tracktag -e TRACKTAG_API_KEY=tt_live_YOUR_KEY -- npx -y tracktag-mcp

```

Replace the placeholder before running the command. Once the assistant restarts or refreshes its MCP connections, check its available tools for TrackTag. If nothing appears, first check that the key is present, Node.js is installed and the configuration is valid. TrackTag’s API and MCP documentation has the current connection details. (tracktag.me)

Claude music analysis: a first track review

Use a specific instruction so the assistant knows what to analyze and how to present the result. For example:

> Find northbound-demo.wav in my local audio files and analyze it with TrackTag. Report the BPM, key, genre and subgenre, moods, instruments, vocal type, song structure and production notes. If there is no steady beat, keep the No BPM result. Do not fill in fields that are missing from the analysis.

The assistant can search for the named file, call TrackTag for the analysis, then present the returned information in the chat. For a quick review, ask for Core. For a delivery where you need the description and broader metadata, ask for Ultra. The analysis level changes the returned fields, not accuracy.

A practical review prompt for a label or library owner might be:

> Analyze northbound-demo.wav with TrackTag using Ultra. Then organize the returned fields into a short intake note with separate sections for sound, vocals, structure and production. Keep the TrackTag values intact and mark any field that was not returned.

That last instruction is useful: it separates measured or analyzed output from the assistant’s own summary. Keep the original result as the source of truth, especially for details such as BPM and key.

Three workflows for musicians, producers and libraries

1. Review a demo without leaving the conversation

A musician or producer can ask Claude to locate a demo by filename, run an Ultra analysis and return the key listening and catalog fields. Use the result to check whether the written description reflects the track, spot unexpected vocal or instrument classifications, and collect a consistent intake note for several collaborators.

The assistant can compare two returned analyses if you ask it to, for example, summarize differences in mood or instrumentation between a main mix and an instrumental. That comparison is the assistant’s interpretation of the TrackTag results. It does not replace listening to both files or confirm that a delivery meets a client’s requirements.

2. Locate a forgotten file and check the credit balance

If you remember only part of a filename, ask the assistant to search your local audio files for it. Then confirm the match before requesting analysis. You can also ask for your remaining TrackTag credits before choosing Core or Ultra. Credit checks and file search do not consume analysis credits, while analyzing a track does. (tracktag.me)

This is a good fit for an occasional task: a producer wants to find a particular demo, a label needs to inspect one delivery, or a library owner wants to answer a metadata question without opening a separate tagging interface.

3. Turn a small review into a catalog handoff

After analysis, ask the assistant to format the returned fields as a concise note for a project record or a message to a colleague. For a structured catalog handoff, use TrackTag Studio to export the whole batch, a selected subset or one file per track. Export formats include CSV or Excel, JSON, XML, Markdown, TXT and PDF. TrackTag Studio batch audio analyzer also supports schema.org JSON-LD in its JSON export.

MCP is most comfortable when a person is directing individual track actions. If you are clearing a large backlog, use TrackTag Studio batch audio analyzer and My Library instead. My Library indexes local filenames and sizes on your machine, shows files as Tagged or Untagged, and provides an Analyze untagged action. Its initial scan does not upload files; a track is sent for analysis only when you choose to analyze it. For a walkthrough of the larger workflow, see the guide to batch processing music files.

What a music MCP server does not automate

An MCP connection is not the same as a background watcher that automatically processes every new file in a folder. It gives an assistant access to specific TrackTag tools during a conversation. You choose what to analyze, and each analysis uses credits. Use the API or Zapier when you need an ongoing intake pipeline, and use Studio or the desktop app when you need to work through a large local folder.

The public API is a better fit for a marketplace that needs to analyze uploads, a label enriching deliveries as they arrive, or a software team showing analysis inside its own product. A request sends a track for analysis and returns results as JSON. Default limits are 10 requests per minute and 2,000 analyses per day per key. The same TrackTag credit balance is used by Studio and the API. (tracktag.me)

Zapier suits no-code handoffs between supported apps. TrackTag’s Zapier app includes an Analysis Finished trigger, an Analyze Track action and a Get Analysis lookup. For example, a Dropbox intake can start an analysis and send the resulting fields to a Google Sheet or Airtable record. The TrackTag Zapier setup guide covers connecting the app and mapping fields.

For a workstation that needs long batches and persistent folder access, TrackTag’s desktop app for Mac and Windows keeps connected folders available without repeated browser permission prompts. Choose the connection that matches the job: MCP for conversational, user-directed analysis; Studio or desktop for hands-on catalog work; Zapier for no-code app handoffs; API for software-driven intake.

Cost and choosing an analysis level

TrackTag offers free credits to start, followed by one-time credit packs rather than a required subscription. Packs are $20 for 50 credits, $89 for 250, $149 for 1,000 and $499 for 5,000. Core uses 1 credit per track; Ultra uses 2. At the 50-credit pack rate, that works out to 50 Core analyses or 25 Ultra analyses for $20. The same balance can be used through MCP, Studio and the API.

For a one-off question, start with Core if its 9 fields are enough. Choose Ultra when you need the full set, including the written description, structure, themes, occasions and production notes. If you later need those additional fields for a Core track, re-analyze it at Ultra for the 1-credit difference.

If you regularly analyze individual tracks while talking through a project with Claude, Cursor or Claude Code, MCP keeps the action close to the work. If the real task is hundreds of files, start in Studio. If each upload should trigger processing without a person opening a chat, set up Zapier or build against the TrackTag public API.

FAQ

What is a music MCP server?

A music MCP server gives an AI assistant access to music-related tools through the Model Context Protocol. TrackTag’s server connects Claude Desktop, Cursor and Claude Code to track analysis, local file search and credit balance checks. (tracktag.me)

Can Claude analyze music files with TrackTag?

Yes. Connect the TrackTag MCP server, provide an API key and ask Claude to find or analyze a specific track. TrackTag returns the analysis, while Claude can organize the results in the format you request.

Does TrackTag MCP analyze a whole folder automatically?

MCP is suited to user-directed actions in a conversation, not unattended processing of every new file. For a folder backlog, connect the folder in TrackTag Studio or use the desktop app; for automated intake, consider Zapier or the public API.

Does using the MCP server cost credits?

Checking credits and searching for a local filename are free. Each completed analysis uses 1 credit for Core or 2 credits for Ultra, and both levels use the same analysis engine.

Tag your whole catalog with AI

BPM, key, genre, moods, instruments and keywords: 30+ fields per track, exported ready for libraries.

Open TrackTag Studio →

← All posts·In-depth guides