How to Analyze Music with AI in Claude: The TrackTag MCP Tutorial (5 Minutes)

You can now analyze music with AI without opening a single app: drop an MP3 into Claude (or Cursor, or any MCP-capable assistant), type "analyze this track", and get back BPM, key, genre, moods, instruments, energy and a full written description, in the chat. This guide sets it up in five minutes using the free, open-source TrackTag MCP server.

What you'll be able to do

• "Analyze sunset-drive.mp3 with TrackTag", full metadata sheet: BPM, musical key, genres and subgenres, moods, emotions, instruments, vocal type, energy/danceability scores and a professional description.
• "Find that demo I downloaded last week and tag it", the server can search your Downloads, Desktop and Music folders by name.
• "How many TrackTag credits do I have left?", balance on demand.

What is MCP, in one paragraph?

MCP (Model Context Protocol) is the open standard that lets AI assistants use external tools, think of it as installing an app for your AI. The TrackTag MCP server runs quietly on your own computer, only while your AI app is open, and connects your assistant to TrackTag's analysis engine.

Setup: five steps, no coding

1. Get an API key. Sign in at tracktag.me/studio → API & MCP tab → Create key. Copy it, it's shown once. New accounts include free credits, and free test keys let you try the wiring without spending anything.
2. Open Claude Desktop's config. Claude menu → Settings → Developer → Edit Config.
3. Paste this block (also copyable from the API & MCP tab), with your key:

{
  "mcpServers": {
    "tracktag": {
      "command": "npx",
      "args": ["-y", "tracktag-mcp"],
      "env": { "TRACKTAG_API_KEY": "tt_live_YOUR_KEY" }
    }
  }
}

4. Restart Claude Desktop fully (Cmd+Q, then reopen, closing the window isn't enough).
5. Drop a track in and ask. Approve the tool the first time, and the analysis appears in the chat.

Using Cursor? Same JSON block in ~/.cursor/mcp.json. Claude Code? One command: claude mcp add tracktag -e TRACKTAG_API_KEY=tt_live_YOUR_KEY , npx -y tracktag-mcp. You'll need Node.js from nodejs.org.

What it costs

Analyses are billed from your TrackTag credit balance, the same one Studio uses: 1 credit per track with the Core model, 2 with Ultra. Credit packs start at $20 for 50 tracks and go as low as $0.10 per track; credits never expire, and failed analyses are refunded automatically. Balance checks and file search are free.

Why this matters for music people

If you tag catalogs for sync libraries, marketplaces or your own releases, your AI assistant just became a tagging workstation. Ask it to analyze a folder of demos while you work, compare two mixes ("which has higher energy?"), or draft store-ready descriptions from real analysis instead of guesswork. And if you build software, the same engine is available as a straight REST API with webhooks.

As far as we know, TrackTag is the first music-tagging platform with a native MCP server, your AI could not do this yesterday. Full documentation, including webhook signatures and test mode, lives at tracktag.me/developers.