About MusicHoarder

What MusicHoarder is

MusicHoarder is an open-source application that turns a folder of badly named, half-tagged audio files into a clean, consistently tagged music library. You point it at a source folder and a destination folder. It fingerprints every track with Chromaprint/AcoustID, so a file called "track03 (1).mp3" is identified by how it sounds rather than by what it is called, asks several metadata providers who agrees on the answer, grades the result, and writes a tidy copy to the destination as plain files: Artist / Year - Album / NN - Track.

The source folder is never modified. MusicHoarder only reads it and writes new copies elsewhere, so a bad match costs you a rebuild and never your originals. Matches the pipeline is not confident about are not guessed into your library — they land in a human review Inbox where you approve, correct or reject them.

How it works

• Scan: the source directory is indexed, and rescanned on a timer so files copied onto the share are picked up without a manual step.
• Fingerprint: each track gets a Chromaprint acoustic fingerprint plus an exact duration.
• Enrich: AcoustID, MusicBrainz, Spotify, Deezer, Apple Music and community trackers are queried in parallel, and a consensus is computed across whichever providers you enabled.
• Grade: an LLM scores the proposed match against the file it came from, so a confident-looking wrong answer still gets flagged.
• Dedupe: duplicates are detected by fingerprint, and the highest-quality copy wins.
• Build: a clean, correctly tagged copy is written to the destination folder, with cover art and synced lyrics where they are available.

Everything runs on your hardware. There is no MusicHoarder cloud, no account to buy and no upload step — the only things that leave your machine are metadata lookups against the public music databases listed above.

Who builds it

MusicHoarder is built and maintained by Jeffrey van den Brink as an open-source side project, in public, under the MIT license. Every line of the API, the web frontend, the Android client and the deployment configuration is in the public repository. Releases are cut automatically from Conventional Commits, and the GitHub Releases page is the canonical changelog.

This site, musichoarder.app, is the project home page. It also hosts a read-only demo account so you can walk through a real library — the Inbox, the match grades, the album pages — before you install anything.

• Source code on GitHub — MIT licensed, issues open
• Releases and changelog
• Self-hosting quickstart — Docker Compose, no repo clone needed
• Machine-readable pricing — it is free

Cost and licensing

MusicHoarder costs nothing. There is no paid tier, no hosted plan and no "contact sales" wall, and there is no plan to add one — it operates on your audio files, which is exactly the kind of data that should stay on your own infrastructure. The software is MIT licensed, so you may use, modify and redistribute it commercially. Some providers it talks to have their own free-tier terms: AcoustID and Spotify need your own API credentials, and MusicBrainz is rate limited.