Audo MCP server and agent plugin

Audo: pro sound without the studio. Audio and video tools for you and your AI assistant: remove noise, enhance voices, transcribe, cut, and more.

This repository packages Audo's hosted MCP server at https://audo.ai/mcp. It uses Streamable HTTP and OAuth. The first time you connect, your assistant asks you to sign in to your Audo account. No API key is required for the OAuth connection.

Install

Agent Plugin clients

This repository contains a portable Agent Plugins manifest, plugin.json, and a root mcp.json pointing to the hosted server. Install the repository using your client's plugin installation workflow. Official marketplace listings are pending; this repository does not imply approval by a marketplace.

Claude Code

claude mcp add --transport http --scope user audo https://audo.ai/mcp

Codex

codex mcp add audo --url https://audo.ai/mcp
codex mcp login audo

Cursor and other MCP clients

For a client's custom MCP configuration, use:

{
  "mcpServers": {
    "audo": {
      "url": "https://audo.ai/mcp"
    }
  }
}

Connect through OAuth when the client prompts. See setup guides for client-specific instructions.

What it does

Speech cleanup: remove_noise reduces background noise while preserving the recorded voice. enhance_voice regenerates supplied speech to restore echo, distortion, muffled recordings, and poor microphones. Both are designed for speech and can remove music or other sounds.
Text and captions: transcribe_audio produces text, word timestamps, speaker labels, and TXT/SRT/VTT/JSON outputs. align_script aligns an existing script to a recording.
Editing: cut_audio, remove_fillers, remove_silence, and join_audio trim recordings, remove filler words or long pauses, and combine files.
Finishing: level_speakers balances loud and quiet voices; mix_tracks layers and ducks tracks; adjust_loudness sets a target loudness; convert_audio changes formats; replace_audio swaps a video's soundtrack.
Music and video: split_audio separates stems; change_speed_pitch changes tempo and pitch; find_key_bpm measures key and tempo; reverse_audio reverses sound; mute_video removes a video's audio.
Analysis: analyze_audio reports media information, loudness, peaks, clipping, and silence.

Use add_file for a direct HTTPS file URL or to start an upload. Supported chat clients also expose Audo's in-chat upload box. Local coding agents can use the returned upload command. Do not pass a local filesystem path as a public URL.

list_files, get_file, get_job, get_transcript, cancel_job, and get_account manage files, status, transcript reading, cancellation, and account information. Processing tools return either finished output or a job to check with get_job.

Example workflows

Remove background noise from an interview and return the cleaned recording.
Balance a quiet guest and loud host, then normalize the result for a podcast.
Transcribe a recording with speaker labels and return the transcript and SRT captions.
Split vocals and instrumental audio, then change the instrumental's speed or pitch.

Free use, limits, and privacy

Tools include a free allowance subject to per-tool daily limits. Full-file processing beyond that allowance requires sufficient existing Audo credits. Current allowances and rates are on Audo's pricing page.

Files and results are deleted after 24 hours. Output download links are temporary; use get_file for a fresh link while the file is retained. Only upload recordings you are permitted to process. See privacy and terms.

Documentation and support

The repository also contains server.json for the official MCP Registry entry ai.audo/audo. The hosted service is operated by Audo AI LLC; its implementation and credentials are not included here.