
MuseScore
STDIOMCP server integrating MuseScore with LLMs for natural language music composition
MCP server integrating MuseScore with LLMs for natural language music composition
This MCP (Model Context Protocol) server integrates MuseScore with LLM clients like Claude Desktop, enabling basic music composition through natural language.
Brief demo here: https://www.canva.com/design/DAGlHeR706g/4UcXvK23I1axWOfm6lSSxQ/edit
The server allows an MCP-connected LLM to:
Clone this repository:
git clone https://github.com/yourusername/musescore-mcp-server.git
cd musescore-mcp-server
Install dependencies, either in a venv or globally:
pip install -r requirements.txt
Install the musescore-mcp-plugin by copying it into your MuseScore plugins dir (on my mac this was ~/Documents/MuseScore4/plugins)
Configure your LLM with the MCP server (server.py). If you are using Claude Desktop and a venv, this can be done by creating a claude_desktop_config file like so:
{ "mcpServers": { "musescore": { "command": "bash", "args": [ "-c", "source /path/to/venv/activate && python3 /path/to/server.py" ] } }
Open musescore and whatever score you want to use. Then, connect the musescore-mcp-plugin and launch it by selecting it from the plugins menu.
Once connected, you can ask Claude questions like:
If you want to add features to the MuseScore plugin and test them, the testClient.html file in this repo may be a useful development tool - simply extend it to call whatever functions you add.