Familiar

Familiar is a self-hosted music streaming service for the library you already own. It analyses every track for what it actually sounds like, not just what it’s tagged as, so you can find music by its sound. You can start a playlist from a single track, album, or artist and get music that sounds like it, not just music filed under the same genre.

Each track gets audio features (BPM, key, energy, danceability) plus CLAP (Contrastive Language-Audio Pretraining) embeddings that support natural-language search over sound. Smart playlists filter on those features with rules. A community analysis cache speeds up processing for common tracks. Artwork fetching, artist and album deduplication, and a feed of new releases from artists already in the library keep a large collection tidy and moving.

Listening happens in real apps: native Familiar Player clients for Mac and iPhone, with offline downloads, CarPlay, and lock-screen controls. The library is also exposed over MCP (Model Context Protocol), so you can point Claude or another AI assistant at your own music and ask it questions or have it build playlists. The server runs on a Mac (Apple Silicon) or anywhere Docker runs: Synology and other NAS boxes, OpenMediaVault, Linux, and Windows. It’s free and MIT-licensed.