Questions

A Mac with Apple Silicon (macOS 14 or later) to run Familiar Server, and the Familiar app on your Mac or iPhone. If you already have a NAS or a Linux box, the server can run there in Docker instead — Install has both. The author runs a 23k-track library on an OpenMediaVault box.

Familiar Server keeps your library and listens to every track to work out how it sounds; Familiar is the app you play it in, on your Mac and your iPhone. They are separate so the server can live wherever suits you — on your Mac, or on a NAS that is always on — and so the app on your phone works the same either way.

To stream from it, yes: the server runs there. Music you have downloaded in the app plays without it, and Familiar Server never keeps your Mac awake — when the Mac sleeps, it waits. If you want your library reachable all the time, run the server on a NAS instead.

Not yet. Familiar Server for Windows, a Windows app and an Android app are planned, after the Mac is finished; there are no dates. Today a Windows PC can run the server in Docker (see Install), and you listen on a Mac or an iPhone.

Familiar plays the files you already own. If you have a folder of MP3s and FLACs going back to 2002, Apple Music either won't take them or will match them to its own catalog versions (sometimes the wrong remaster, sometimes gone); Spotify plays only its own catalog, and its Local Files work from a desktop to a phone on the same network. Neither listens to your files to find music by how it sounds. Familiar does, on your own machine, and leaves the files alone.

No. Familiar holds no API key and calls no model; search, playlists, similarity and the Music Map all work on their own. If you use an AI assistant that speaks MCP — Claude Desktop, say — you can connect it to ask for music in words, and whatever that costs is between you and them.

Your audio files and your listening history don't. To fill in artwork, artist pictures, lyrics and metadata, the server sends the names it is looking up — artists, albums, titles — to public services such as MusicBrainz and Wikipedia; the community cache gets a one-way hash of each track's audio fingerprint, from which no library can be reconstructed. Scrobbling to Last.fm and connecting an AI assistant happen only if you set them up. The privacy page lists every one.

On a large library, days rather than hours, and on a Mac it pauses while you are on battery. You can listen while it works. The opt-in community cache skips tracks that have already been analysed elsewhere.

Yes — see Listening away from home. Tailscale is the easiest way: it puts your phone and your server on one private network, so the app reaches the server on cellular as it does at home.

Cloudflare Tunnel if you want a stable public hostname routed through Cloudflare without opening a port. WireGuard if you would rather run your own VPN and use the LAN address. Or any VPN you already run: anything that puts your phone on the same network as the server will do. A reverse proxy — nginx, Caddy, Traefik — does not make Familiar safe to publish: with HTTPS it keeps the sign-in private on the way, but the music needs no sign-in (why). What you should not do is port-forward 4400 to the open internet.

Yes, and it is where listening on a phone is meant to happen — Familiar is on the App Store, with a Mac app alongside it. It plays from a Familiar server, so you need one running before the app is useful.