# Sonicmatch

> Watches the footage, then returns license-safe music, a hook window, and a ducking spec.

- **Category:** Music & Score
- **Vendor:** Community (js713-lab) (community)
- **Pricing:** free
- **Verified:** 2026-09-23

## What it does

Sonicmatch is music supervision shaped for an agent: ingest a cut, analyse mood, energy curve, speech coverage and scene cuts, then get back a ranked shortlist of tracks with the license printed on every row, a 12–20 second hook in/out, an ffmpeg ducking recipe, and a beat-snapped cut list. It reads the picture rather than the script, which is the difference between a music search and a music choice. The intended production use is pointing it at an Artlist or Epidemic library you already pay for, via a JSON manifest, so the scoring logic runs over a catalogue you are actually cleared to deliver.

## Connect

- Claude Code: `claude mcp add sonicmatch -- uvx --from git+https://github.com/js713-lab/sonic-match-mcp.git sonicmatch-mcp`
- Local: `uvx --from git+https://github.com/js713-lab/sonic-match-mcp.git sonicmatch-mcp — or pip install git+https://github.com/js713-lab/sonic-match-mcp.git then run sonicmatch-mcp. Needs Python 3.10+ and ffmpeg/ffprobe on PATH; add --http --port 8765 for Streamable HTTP instead of stdio.`

**Auth:** none — No key required. Optional: GEMINI_API_KEY for video understanding, JAMENDO_CLIENT_ID and FREESOUND_API_KEY for extra catalogues, SONICMATCH_LIBRARY_PATH / EPIDEMIC_LIBRARY_PATH / ARTLIST_LIBRARY_PATH for a library you already licensed

## Capabilities

[video-analysis](https://mcp.film/capabilities/video-analysis.md) · [music-search](https://mcp.film/capabilities/music-search.md) · [music-licensing](https://mcp.film/capabilities/music-licensing.md) · `beat-detection` · [scene-detection](https://mcp.film/capabilities/scene-detection.md) · `ducking` · `mix-spec`

## Sample tools

`status` · `ingest_video` · `analyze_video_music` · `recommend_bgm` · `search_music` · `get_track` · `preview_mix` · `export_mix_spec` · `suggest_cuts` · `generate_bed` · `save_brand_kit` · `analyze_batch`

## Field notes

Young and thin, listed for the shape of the idea rather than the catalogue behind it. Checked 23 Sep 2026: MIT, first public release 10 Sep 2026, one v0.2.0 tag, last commit 16 Sep, a single maintainer, 94 stars against zero forks — and published to the official MCP registry as io.github.js713-lab/sonicmatch-mcp with an .mcpb bundle. Not on PyPI, so every install path is a git checkout; the twelve tools above were read off the source, not just the README. The built-in catalogue is 22 seed tracks whose preview and download URLs are null on purpose, so out of the box recommend_bgm hands you rows you cannot fetch and preview_mix synthesises a demo bed instead of real music — the useful configuration is your own licensed Artlist or Epidemic manifest, or the optional Jamendo and Freesound keys. Platform URL ingest via yt-dlp ships off by default for ToS reasons; leave it off and feed it local files. The licensing discipline is the best thing here — non-commercial tracks are never auto-recommended and generated beds are excluded from recommendations — but a CC label is not a Content ID waiver, and CC-BY still needs attribution in your credits. For a cleared catalogue with the rights already settled, the Epidemic Sound entry is the grown-up option; Sonicmatch is the analysis layer you would put in front of it.

## Verification

- Status: Freshly checked (checked today)
- Added: 2026-09-23
- Last verified: 2026-09-23
- Machine JSON: https://mcp.film/api/mcps/sonicmatch.json

## Links

- Docs: https://github.com/js713-lab/sonic-match-mcp?utm_source=mcp.film&utm_medium=referral&utm_campaign=directory&utm_content=sonicmatch%3Adocs#readme
- Repo: https://github.com/js713-lab/sonic-match-mcp

---

Structured data: https://mcp.film/api/mcps/sonicmatch.json · Directory: https://mcp.film
Canonical: https://mcp.film/mcps/sonicmatch/
Hosted directory MCP: https://mcp.film/mcp