MuseScore MCP Server

What is MuseScore MCP Server?

This MCP server integrates MuseScore with LLM clients like Claude Desktop, enabling basic music composition through natural language commands. It is intended for musicians and composers who want to use AI to interact with an open MuseScore score programmatically.

How to use MuseScore MCP Server?

Install Python 3.9+ and MuseScore 3 or 4, then clone the repository and install dependencies. Copy the musescore-mcp-plugin into your MuseScore plugins directory, then configure your LLM client (e.g., Claude Desktop) to run server.py via a JSON config. Open MuseScore, load a score, and launch the plugin from the Plugins menu. Afterwards, you can issue natural language queries such as "Compose a simple melody" or "Help me come up with possible chords for the melody."

Key features of MuseScore MCP Server

• Connect to MuseScore and manipulate an open score
• Add notes and rests to the score
• Select and delete elements
• Create tuplets (e.g., triplets)
• Undo recent changes
• Move between staves within the score

Use cases of MuseScore MCP Server

• Ask an LLM to suggest harmonies for an existing melody in your score
• Have the LLM compose a short melody and present multiple harmonization options
• Extend an existing piece by generating measures that match the original style
• Automate repetitive editing tasks like adding or deleting notes via text prompts

FAQ from MuseScore MCP Server

What are the runtime requirements?

Python 3.9 or later, MuseScore 3 or 4 installed locally, and the MCP Python SDK (installed via pip install -r requirements.txt).

How do I install the MuseScore plugin?

Copy the musescore-mcp-plugin folder into your MuseScore plugins directory (e.g., ~/Documents/MuseScore4/plugins on macOS). Then select the plugin from MuseScore’s Plugins menu and launch it.

Can the server handle multiple voices within a staff?

No. The current limitation is that it cannot understand or navigate between multiple voices within a single staff.

Where does the data live?

The server operates on the local MuseScore file you have open; no data is sent to external servers beyond the LLM client you configure (e.g., Claude Desktop).

Which transport does the server use?

The server communicates via the Model Context Protocol (MCP) over standard input/output (stdio).