{
  "slug": "audiolab",
  "name": "AudioLab MCP",
  "vendor": "AudioLaunch",
  "official": true,
  "category": "sound-design",
  "featured": false,
  "tagline": "Loudness, true-peak and voice metering that grades a mix against real delivery specs.",
  "description": "AudioLab is the meter most generative sound pipelines are missing: ten analysis tools over a hosted API returning integrated LUFS to EBU R128 / ITU BS.1770-4, true-peak in dBTP, loudness range, crest factor, stereo correlation, spectrum, and voice diagnostics down to speaking rate, SNR, room echo, sibilance and clipping risk. check_target grades a file against a named delivery spec — ebu-broadcast, atsc-broadcast, podcast or the streaming platforms — and returns per-metric deltas plus the ffmpeg loudnorm command that would fix it. Nothing here generates or processes audio; it tells an agent whether the mix passes before the deliverable ships, which is the step between a finished cut and a rejected master.",
  "capabilities": [
    "loudness-analysis",
    "true-peak",
    "delivery-qc",
    "voice-analysis",
    "spectrum-analysis",
    "speech-segmentation",
    "batch-analysis"
  ],
  "tools_sample": [
    "analyze_loudness",
    "check_target",
    "analyze_timeseries",
    "get_spectrum",
    "analyze_voice",
    "get_speech_segments",
    "index_signal",
    "analyze_profile",
    "compare_loudness",
    "analyze_batch"
  ],
  "install": {
    "claude_code": "claude mcp add audiolab -e AUDIOLAB_API_KEY=YOUR_KEY -- npx -y @audiolabtools/mcp-server",
    "remote_url": "https://audiolab.tools/mcp",
    "remote_headers": [
      "Authorization"
    ],
    "stdio_command": "npx -y @audiolabtools/mcp-server"
  },
  "auth": {
    "type": "api_key",
    "env_var": "AUDIOLAB_API_KEY",
    "key_url": "https://audiolab.tools/account"
  },
  "pricing": "freemium",
  "links": {
    "repo": "https://github.com/Audio-Launch/audiolab-mcp-server",
    "docs": "https://audiolab.tools/api",
    "site": "https://audiolab.tools"
  },
  "added": "2026-09-23",
  "verified": "2026-09-23",
  "notes": "Verified 23 Sep 2026: MIT, npm @audiolabtools/mcp-server at 0.4.1 published the day before, nine releases since 3 Jul 2026, and listed in the official MCP registry. Take the stdio form if your audio is on disk — a local `path` works only in the npx server; the hosted endpoint at https://audiolab.tools/mcp (401 unauthenticated, `Authorization: Bearer` rather than OAuth) accepts public URLs only. Either way it is a hosted analyser: a URL is fetched server-side, a local file is uploaded, measured and deleted, and only numbers come back, so if the audio cannot leave the building do not use it. Free tier is 1,000 calls and 2 GB a month with no card, but the seven named analyze_profile lenses (voice, master, provenance, dataset, environment, broadcast, loop) are plan-gated; free keys get `basic`. Limits that bite a feature: 50 MB per local file, loudness routes stream to roughly three hours, voice and signal routes cap near seven minutes, one file per call, and analyze_batch tops out at 20 sources while metering each as its own call. It measures and never fixes — the fix is an ffmpeg loudnorm string you run yourself, or the Auphonic entry for actual processing. Operated by AudioLaunch (KVK 54871204, EU-hosted), built by Nathan Renting; the GitHub repo is a generated mirror of an internal monorepo, so read the npm cadence rather than the star count."
}