@cynosure-mcp/music-tagger

MCP server for automatically identifying and tagging music files using audio fingerprinting (AcoustID + Chromaprint) and metadata from MusicBrainz.

Tools

| Tool | Description | | ----------- | ---------------------------------------------------------------------- | | read_tags | Read current embedded tags from a file or all music files in a folder | | tag_music | Fingerprint and write accurate tags (+ cover art) to one or more files |

Requirements

API Key

Register for a free AcoustID API key at https://acoustid.org/api-key and pass it as ACOUSTID_API_KEY.

fpcalc (optional)

The MCP is fully self-contained — audio is decoded using the bundled ffmpeg-static binary and fingerprinted via rusty-chromaprint-wasm (a WASM build of Chromaprint; no native system libraries needed).

If fpcalc (Chromaprint CLI) is already installed on the system it will be used automatically as a faster alternative.

| Platform | Install command | | -------- | ------------------------------------------------ | | Linux | sudo apt install libchromaprint-tools | | macOS | brew install chromaprint | | Windows | Download from https://acoustid.org/chromaprint |

Configuration

| Variable | Required | Description | | ------------------ | -------- | ------------------------------------------------ | | ACOUSTID_API_KEY | Yes | Free API key from https://acoustid.org/api-key |

Supported Formats

MP3, FLAC, OGG Vorbis, M4A/AAC, WAV, WMA, Opus, APE, Musepack

Tags Written

• Title, Artist, Album Artist, Album
• Year, Track number / total, Disc number / total
• Genres (up to 3, from MusicBrainz community votes)
• Cover art (250×250 JPEG from Cover Art Archive, embedded)

MCP Config

{
  "mcpServers": {
    "music-tagger": {
      "command": "npx",
      "args": ["@cynosure-mcp/music-tagger"],
      "env": {
        "ACOUSTID_API_KEY": "your_key_here"
      }
    }
  }
}

How It Works

1. Decode — The first 120 s of the audio file is decoded to raw PCM using the bundled ffmpeg-static binary
2. Fingerprint — rusty-chromaprint-wasm (WASM, no system deps) generates an AcoustID-compatible fingerprint; fpcalc is used instead when available
3. Identify — The fingerprint is submitted to the AcoustID API, which returns a MusicBrainz Recording ID
4. Fetch metadata — Full recording metadata (title, artists, releases, genres) is fetched from MusicBrainz
5. Cover art — Album artwork is fetched from the Cover Art Archive (250×250)
6. Write — All tags are embedded into the file using TagLib (via node-taglib-sharp)

MusicBrainz requests are rate-limited to ≤1 req/s as required by their API policy.

License

MIT