mcp-lyricscom

lyrics.com is a large public catalogue of song lyrics. It files a song under its title, its artist, the album it appeared on and the year, and it holds the words themselves. Its search reaches inside those words.

This server connects a chat client to that catalogue. You can search for a song by a line you remember, search by title and artist, and read the words of one song, a slice at a time, with the words you were looking for located in the text. It needs no API key and no account.

Version française

Install

One-click install

Claude Code

claude mcp add lyricscom -- npx -y mcp-lyricscom

Claude Desktop, Cursor, and any client using the standard config format

{
  "mcpServers": {
    "lyricscom": {
      "command": "npx",
      "args": ["-y", "mcp-lyricscom"]
    }
  }
}

Node 24 or later is required, and no environment variable has to be set.

With Docker

{
  "mcpServers": {
    "lyricscom": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "ghcr.io/smeet666/mcp-lyricscom:2.0.1"]
    }
  }
}

-i keeps stdin open, which is where the protocol travels, and -t is left out because a TTY rewrites the stream. The container needs outbound HTTPS to www.lyrics.com, and nothing else: no volume, no port, no credential.

Bundle, without npm

Download mcp-lyricscom-2.0.1.mcpb from the latest release and open it. A client that supports MCP bundles installs it on its own, with no npm and no configuration file to edit. The bundle carries its dependencies, so nothing is fetched at install time.

What you can ask

• "Which song goes 'I've got a hand for you'?"
• "Find me the lyrics of Wichita Lineman by Glen Campbell."
• "Read me the second half of those words."
• "Where does the word 'lineman' appear in that song?"
• "What albums is that song on?"

The ordinary path runs from a search to a reading: a row carries an id, and get_lyrics takes that id.

Tools

search_lyrics

Finds a song from words inside its lyrics. The site ranks loosely, so a match is checked before it is served.

verify decides what a row is worth. snippet checks the excerpt the site already returned and costs nothing. full fetches up to five song pages and checks the complete words, which is slow and can bring on rate limiting. none serves what the site ranked, unchecked.

In return: rows carrying id, which get_lyrics takes; title; artist; album and year, null where the catalogue states none; source_url; and excerpt, the matching line. raw_result_count is what the site returned and filtered_out how many rows the check removed, so the two together say how loose the ranking was. has_more and next_page continue.

search_songs

Finds songs by title, narrowed by artist.

In return: the rows search_lyrics returns, with artist_filter echoing what was asked for and filtered_out counting what the artist restriction removed. strict keeps the artists whose name matches as written; loose accepts a name written differently.

get_lyrics

Reads the words of one song. Long lyrics are served a slice at a time.

In return: status, reading ok or no_lyrics for a page the site holds without words; title, artist and source_url; and lyrics, the slice itself. The reading is described by total_chars, returned_chars, offset, next_offset and truncated: pass next_offset back to read on, and null there means the end. line_count counts the lines of the slice, and highlight answers for each word whether it was found and on which line_number, which is null when it was not.

Configuration

Every variable is optional. Set them in the env block of your client config.

A value outside its range falls back to the default, and the reason is written to stderr.