Identify any song from a link or an audio file, then read its BPM, musical key and Camelot code — from your terminal or your code.
Official clients for Song Finder, a free online song finder that identifies music from a file, a microphone recording or a link. No API key. No account. No rate card.
npx songfinder identify "https://www.youtube.com/watch?v=..."
npx songfinder analyze "strobe deadmau5"I Remember (Strobelight Edit) — deadmau5
Strobelite Seduction (2013)
ISRC: USUS10800096
Tempo: 128 BPM
Key: B Minor (Camelot 10A, Open Key 3m)
Mixes with: 10B, 9A, 11A
Energy: 59%
Danceability: 66%
Valence: 49%
Loudness: -7.2 dB
(Real output, not a mock-up.)
npm install -g songfinder # CLI on your PATH
npm install songfinder # as a libraryNode.js 20+. Zero runtime dependencies.
songfinder identify <url|file> [--start <seconds>]
songfinder search <query>
songfinder analyze <query|ISRC>
songfinder similar <query|ISRC> [--limit <n>] [--harmonic]
Every command that names a track accepts either an ISRC or a plain search phrase — you never have to look a code up by hand first. Add --json anywhere to get the raw API response instead of formatted text.
# what is playing in this video
songfinder identify "https://www.tiktok.com/@user/video/123..."
# what is this file on disk
songfinder identify ~/Music/unknown.mp3
# sample 90 seconds in, when the opening is silence or an intro
songfinder identify "https://youtu.be/..." --start 90
# five tracks that mix harmonically into this one
songfinder similar "strobe deadmau5" --harmonic --limit 5import { SongFinder, tempoDisagrees } from 'songfinder';
const sf = new SongFinder();
const [track] = await sf.search('blinding lights weeknd');
const detail = await sf.detail(track.isrc);
console.log(detail.features?.tempo); // 171.005
console.log(detail.features?.key?.camelot); // "3B"
console.log(detail.features?.key?.compatible); // ["3A", "2B", "4B"]
if (tempoDisagrees(detail)) {
// a second provider read this at half or double time — see below
}| Method | Returns |
|---|---|
search(query) |
Up to 10 catalogue candidates with ISRCs |
detail(isrc) |
Tempo, key, Camelot code, energy/danceability/valence, genre |
similar(isrc, { limit, harmonic }) |
Neighbouring tracks, each with its own tempo and key |
identifyUrl(url, startSeconds?) |
Recognition from a page or media URL |
identifyFile(bytes, name, mime) |
Recognition from an audio buffer (max 10MB) |
resolveIsrc(title, artist) |
Bridges recognition output to the analysis methods |
Everything is fully typed. SongFinderError carries the HTTP status when there was one.
Half-time and double-time tempo. detail() returns a second provider's reading in tempoCrossCheck. When the two disagree by more than 3 BPM, one of them counted the groove at half speed — an 87/174 pair is the same track. tempoDisagrees(detail) tells you; the CLI prints a warning. Reporting the primary figure alone is how tempo data most often misleads people.
Recognition returns no ISRC. Every analysis endpoint is keyed by ISRC, so the chain would dead-end right after the interesting part. resolveIsrc() closes that gap with one catalogue search, and never throws — a lookup failure must not sink a successful match.
Recognition is rate-limited per IP because it spends paid third-party quota. The analysis endpoints allow roughly one call per 1–5 seconds depending on how far they fan out upstream. A 429 means you went too fast, not that the track is missing.
The credited artist is not always the original artist. Widely re-uploaded tracks match white-label catalogue entries, so a famous song can come back credited to a label nobody has heard of. The title is still right — search that title to find the original release.
Audio you identify is uploaded to songfinder.dev and is not retained. Analysis calls send only a text query or an ISRC — no audio.
Coverage is uneven for long-tail releases. Many have a tempo but no key, or no analysis at all. Missing fields come back null rather than being invented.
| songfinder-mcp | MCP server — the same capabilities inside Claude, Cursor, Windsurf or Zed |
| songfinder-skills | Claude Code Agent Skills, no install beyond curl |
| Song Finder — free online song finder | The web app: identify by file, microphone or link, plus BPM/key detection, a Camelot wheel, stem separation, audio trimming and more |
| Song finder browser extensions | Chrome, Edge, Firefox and a userscript |
js/ npm package "songfinder" — client + CLI
assets/ README artwork
cd js
pnpm install
pnpm build
node dist/cli.js analyze "strobe deadmau5"MIT