CDP MCP Server

A Model Context Protocol (MCP) server that provides direct access to the Composers' Desktop Project (CDP) sound transformation programs. This server offers an ultra-rigid workflow with zero interpretation, exposing CDP's raw functionality through simple, reliable tools.

CDP MCP Server v7 implements a minimalist approach to CDP integration:

- Direct execution- No command parsing or interpretation
- Raw usage text- See exactly what CDP shows
- Simple tools- Just 6 core functions
- Data file support- Create parameter files for complex operations

- 🎵Full CDP Access- Execute any CDP program with complete control
- 📚Program Discovery- List all available CDP programs by category
- 📖Usage Information- Get raw usage text directly from CDP
- 📄Data File Creation- Create parameter files required by CDP programs
- 🎛️Spectral Preparation- Helper for PVOC analysis
- 📊Sound Analysis- Get basic properties of audio files

- macOS, Linux, or Windows
- Python 3.8 or higher
- CDP (Composers' Desktop Project) installed
- Download CDP from theofficial website
- Install CDP following the platform-specific instructions
- Set theCDP_PATHenvironment variable to your CDP programs directory:

export CDP_PATH="/path/to/cdp/programs"

git clone https://github.com/DavidPiazza/CDP_MCP.git cd CDP_MCP

export CDP_PATH="/Users/yourname/cdpr8/_cdp/_cdprogs" # macOS example

{ "mcpServers": { "cdp": { "command": "python", "args": ["/path/to/CDP_MCP_v7.py"], "env": { "CDP_PATH": "/path/to/cdp/programs" } } } }

list_cdp_programs() # Returns categorized list of all CDP programs

get_cdp_usage('blur') # Returns exact CDP usage text

execute_cdp(['blur', 'blur', 'input.ana', 'output.ana', '50']) # Direct execution with no interpretation

# Check usage get_cdp_usage('stretch') # Execute execute_cdp(['stretch', 'time', '1', 'input.ana', 'output.ana', '2.0'])

# Note the double syntax for blur execute_cdp(['blur', 'blur', 'input.ana', 'output.ana', '50'])

execute_cdp(['modify', 'brassage', '4', 'input.wav', 'output.wav', '0.02', '-0.5', '-r200'])

# Create data file for tesselate create_data_file('tess_data.txt', '5 5 5 5\n0.0 0.1 0.2 0.3') # Execute with data file execute_cdp(['tesselate', 'tesselate', '1', 'in1.wav', 'in2.wav', 'in3.wav', 'in4.wav', 'output.wav', '0.5', 'tess_data.txt'])

Lists all available CDP programs organized by category (Spectral Processing, Time Domain, Synthesis, etc.)

Returns raw usage information for a CDP program by running it without arguments.

Executes a CDP command given as an array of strings. No parsing or interpretation.

Creates text data files required by certain CDP programs.

prepare_spectral(input_file, output_file, window_size=2048)

Helper function to perform PVOC analysis for spectral processing.

Returns basic properties of a sound file (duration, sample rate, channels, etc.)

The server follows an ultra-rigid design philosophy:

- No command parsing- Commands are arrays, not strings
- No parameter validation- CDP handles all validation
- No output interpretation- Raw CDP output is returned
- Minimal abstraction- Direct CDP access only

- Always check usage withget_cdp_usage()before executing
- CDP may return non-zero exit codes even on success
- Check if output files exist to verify successful execution
- Use exact command arrays - the server does no interpretation
- Create data files when CDP usage mentions DATAFILE requirements

EnsureCDP_PATHenvironment variable points to your CDP programs directory.

The server automatically handles x86_64 emulation on Apple Silicon Macs usingarch -x86_64.
- Check the exact usage withget_cdp_usage()
- Verify all file paths exist
- Ensure proper command array format
- Check CDP's stderr output for specific errors

MIT License - See LICENSE file for details

- Composers' Desktop Projectfor the amazing sound transformation tools
- Model Context Protocolfor the MCP framework

Build and deploy full-stack Next.js apps with 98 tools for React, AWS, and MongoDB