CleanAudio MCP documentation

Hosted endpoint, OAuth sign-in, tool reference, and example prompts for Claude, Cursor, ChatGPT, and VS Code.

MCP

Copy the CleanAudio connector URL

You'll paste this URL into Claude in the next step

Go to Claude → Customize

In Claude desktop or claude.ai, go to Customize → Connectors. Name it CleanAudio and paste the URL

Open Claude Customize

Connect, sign in and start

Start cleaning

MCP endpoint

CleanAudio is a hosted MCP server for audio editing. Point your client at this URL. Sign in with your CleanAudio account. No self-hosting and no FFmpeg install for the remote path.

https://mcp.cleanaudio.app/mcp

Claude: Settings → Connectors → add custom connector
Cursor: remote MCP server URL in settings
ChatGPT and other MCP clients that support remote HTTP MCP
VS Code / Claude Code: remote MCP config with the same URL

Authentication

Paste the connector URL into your MCP client.
Start a connection. Your browser opens CleanAudio.
Sign in and approve access for the client.
Your agent can call tools as your user. Usage follows your plan credits.

Think of CleanAudio MCP as the Stripe-style layer for audio processing: hosted auth, usage-based credits, and a predictable tool API instead of a self-hosted FFmpeg wrapper.

Available tools

Clear tool descriptions help models pick the right action. Each tool below matches the live CleanAudio MCP server.

clean_noise

Remove background noise and improve speech clarity. Accepts a file URL or base64 audio.

Parameters

audio_url (optional URL)
audio_base64 (optional)
filename (optional)
strength (0-100, default 70)

Example request

{
  "audio_url": "https://example.com/episode.mp3",
  "strength": 70
}

Example response

{
  "success": true,
  "audio_url": "https://...",
  "filename": "episode-cleaned.mp3",
  "strength": 70
}

cut_audio

Trim an audio file between start and end times in seconds. Returns the clipped audio.

Parameters

audio_url or audio_base64
start (seconds, >= 0)
end (seconds, > start)
filename (optional)

Example request

{
  "audio_url": "https://example.com/take.wav",
  "start": 0,
  "end": 30
}

Example response

{
  "success": true,
  "audio_url": "https://...",
  "start": 0,
  "end": 30
}

join_audio

Concatenate 2-10 audio file URLs into one MP3 in order.

Parameters

audio_urls (array of 2-10 HTTPS URLs)

Example request

{
  "audio_urls": [
    "https://example.com/part-a.mp3",
    "https://example.com/part-b.mp3"
  ]
}

Example response

{
  "success": true,
  "audio_url": "https://...",
  "parts": 2
}

remove_silence

Remove long silences and dead air to tighten pacing on voice recordings.

Parameters

audio_url or audio_base64
threshold (dB, default -50)
filename (optional)

Example request

{
  "audio_url": "https://example.com/interview.mp3",
  "threshold": -50
}

Example response

{
  "success": true,
  "audio_url": "https://...",
  "threshold_db": -50
}

text_to_speech

Convert text to speech audio with Edge TTS voices.

Parameters

text (1-5000 characters)
voice (default "en-US-JennyNeural")

Example request

{
  "text": "Welcome to this week's episode.",
  "voice": "en-US-JennyNeural"
}

Example response

{
  "success": true,
  "audio_url": "https://...",
  "voice": "en-US-JennyNeural",
  "characters": 32
}

transcribe

Transcribe speech in an audio file to text.

Parameters

audio_url or audio_base64
language (default "English")
filename (optional)

Example request

{
  "audio_url": "https://example.com/notes.m4a",
  "language": "English"
}

Example response

{
  "success": true,
  "transcription": "...",
  "language": "English",
  "word_count": 420
}

Example prompts

“Trim the first 30 seconds of this recording”
“Remove silence and dead air from this interview”
“Clean background noise from this podcast clip”
“Join these two MP3 files in order”
“Transcribe this voice note to text”
“Turn this script into speech audio”

MCP overview Audio editing MCP FFmpeg alternative

MCP docs FAQ

Use the remote connector URL https://mcp.cleanaudio.app/mcp. Paste it into Claude Connectors, Cursor remote MCP settings, ChatGPT MCP, or another MCP-compatible client.

CleanAudio uses OAuth account sign-in. After you add the connector, your browser opens CleanAudio. Sign in and approve access. There is no API key to paste for the hosted connector.

No. CleanAudio is a hosted MCP service. Processing runs on CleanAudio infrastructure with your plan credits. You do not install FFmpeg or run a local server for the remote connector.

MCP calls use your CleanAudio plan the same way the web app does. Starter is $9/month and Unlimited is $20/month. MCP access is included. There is no separate per-call MCP surcharge.