MCP Bideetmusique

mcp-bideetmusique

Bide & Musique is a French web radio and the collection of records it plays, catalogued by hand by the volunteer association that runs it. Its subject is the forgotten song: the flop, the novelty record, the television theme, the single nobody reissued. A record's page carries the performer, the writers and composers, the year, the label and its catalogue reference, the sleeve, the duration, the station's own chart placings, and the words when a listener transcribed them.

This server connects a chat client to that collection. You can search it along one axis at a time, by performer, title, writer, lyrics, label or year; read one record in full; read what the site holds about an artist with their discography; draw a record at random; and see what has just been catalogued. It needs no API key and no account.

_Version française_

Install

One-click install

Claude Code

claude mcp add bideetmusique -- npx -y mcp-bideetmusique

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

{

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

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

With Docker

{

"mcpServers": { "bideetmusique": { "command": "docker", "args": ["run", "-i", "--rm", "ghcr.io/smeet666/mcp-bideetmusique:2.1.0"] } } }

-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.bide-et-musique.com, and nothing else: no volume, no port, no credential.

Bundle, without npm

Download mcp-bideetmusique-2.1.0.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

• « Qu'est-ce que Bide & Musique a de Jacques Dutronc ? »
• "Which songs mention a submarine in their words?"
• "Read me that record: the label, the year, who wrote it."
• "Play me something at random from the collection."
• "What has been catalogued lately?"

The ordinary path runs from a search to a record: a row carries a song_id, and get_song reads it.

Tools

search_songs

Searches along one axis, which has to be named: each asks a different question, and an answer on one says nothing about the others. A name that returns nothing as a performer may still be a title.

The year axis takes one four-digit year and nothing else: the site drops any other word there instead of filtering on it. Several keywords are combined with AND and each is matched inside words, so every extra word narrows the search, and a quoted phrase returns nothing whatever the site's own form offers.

In return: rows carrying song_id, which get_song takes; title; url; and the artist. total_matches is the number the site prints, counting matching songs across every page, so it exceeds the rows of one page. page_served is the page the site actually served, since it answers a page past the last one with the last page and no error: read it rather than assuming the page asked for. page_count and has_more_pages are null where the site printed neither.

get_song

Reads one record in full.

In return: title, artist, credited_performer, year, writers, duration with the text exactly as the record prints it and its precision, labels with one entry per label, catalogue_reference, presentation, sleeve_credits, image_url, added_on as an ISO date, the station's chart placings, favourites and comments. The title, the artist and the duration are always stated; the rest is absent on some records and comes back null rather than guessed, and a counter the page prints nothing for is unknown rather than zero. lyrics.available says whether the page carries a transcription, and lyrics.text holds the words with the transcriber who wrote them down.

get_artist

Reads what the collection holds about one artist, with the records it has of theirs.

In return: name, aliases, surname, first_name, nationality, birth_date, presentation, see_also, links and photo_url, each null or empty where the page states nothing. discography carries their records, each with its song_id for get_song, its title, url, year and how often the station programmed it. discography_count counts the records served and songs_on_page the records the page held, so a truncated discography is visible.

get_random_song

Draws a record nobody chose, over the ids the site serves. It is for browsing the collection when no particular song is being asked about.

In return: the record get_song returns.

list_new_songs

Reads what the collection has just catalogued.

In return: rows carrying song_id, title read off the line the feed publishes, artist_name, listed_as, url and published_at. entries_in_feed is the number of entries the feed holds, which is a fixed number and says nothing about how many records the collection holds.

Configuration

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

A value ou