Audiolla — Claude Code MCP server: what it does, how to install
A self-hosted MCP server running inside Docker for local audio tasks. It gives coding assistants tools for stem splitting and speech cleanup without external cloud calls.
Open on GitHubWhat Audiolla does
Audiolla packages thirty audio engines into a single Docker image exposed through an HTTP API and an MCP endpoint. It includes Demucs for stem isolation, basic-pitch for polyphonic note detection, and DeepFilterNet with pyannote for vocal enhancement. The container accepts an audio file and returns extracted stems, analytical JSON data, or rendered MIDI files.
The container isolates system dependencies and avoids Python library clashes across models. Heavy tasks run in the background. They report completion via async_job=true and webhook callbacks. Running this server requires a local Docker engine with enough resources to pull and run psyb0t/audiolla.
What it helps with
- Separate mixed tracks into individual stems using Demucs
- Transcribe recorded instruments into MIDI sequences
- Remove room reverberation, background noise, and electrical hum
- Measure BPM, musical key, and integrated LUFS loudness
- Label speaker segments in conversation files with pyannote
Example prompts for the agent
- Split the file
track.mp3into stems and output the vocal track - Convert the guitar recording in
solo.wavinto a MIDI file - Clean the background noise from
speech.wavand normalize its LUFS level
How to install Audiolla
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Audiolla
How to install Audiolla in Claude Code?
Start the container via docker run using the psyb0t/audiolla image, then add the exposed MCP endpoint to your Claude Code configuration. The required models and DSP libraries are prepackaged inside.
What does Audiolla do?
It provides audio engineering tools to your assistant, covering stem isolation, DSP effects, loudness metering, and acoustic tagging. Every operation runs on your own hardware without external subscriptions.
Which processing tools are bundled in the container?
The setup bundles Demucs, basic-pitch, DeepFilterNet, pyannote, and DSP libraries such as librosa and Sox. They are accessible through one unified interface.
