Pd
FreeNot checkedA Model Context Protocol (MCP) server for Pure Data, an open-source visual programming language and patchable environment for real-time computer music.
About
A Model Context Protocol (MCP) server for Pure Data, an open-source visual programming language and patchable environment for real-time computer music.
README
This project provides integration between Claude AI and Pure Data through the Model Context Protocol (MCP). With this integration, Claude can dynamically create, modify, and control Pure Data patches through natural language.
Overview
The system consists of three core components:
- MCP Server (
mcp_server.py): Implements the Model Context Protocol interface for Claude - OSC Daemon (
osc_daemon.py): Handles OSC communication with Pure Data - Pure Data Patch (
example_patch.pd): A dynamic patching solution that receives OSC messages
Features
- Dynamic Object Creation: Create any Pure Data object on demand through Claude
- Connection Management: Connect objects together to build complex signal flows
- Workspace Management: Clear all objects and reset state with
clear_workspace - DSP Control: Start and stop audio processing remotely
- Parameter Control: Modify parameters of objects in real-time
- Global Object Tracking: Reliable index-based connection system
- Error Handling: Robust error detection and reporting
- Debugging Tools: Comprehensive logging system
Recent Updates
Installation
Prerequisites
- Python 3.7+
- Pure Data (vanilla) 0.51+
- Required Python packages:
python-oscfastmcpjsonschema
- uv package manager
Install uv
On macOS and Linux:
curl -LsSf https://astral.sh/uv/install.sh | sh
On Windows (PowerShell):
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
Learn more: astral-sh/uv
🛠️ Usage
Clone the repository
git clone https://github.com/nikmaniatis/Pd-MCP-Server.git
Integration with Claude Desktop
- Open:
Claude > Settings > Developer > Edit Config > claude_desktop_config.json
- Add the following block:
{
"mcpServers": {
"Pure Data MCP Server": {
"command": "uv",
"args": [
"--directory",
"PATH_TO_PD_MCP_SERVER",
"run",
"main.py"
],
"env": {
"PD_OSC_HOST": "127.0.0.1",
"PD_OSC_PORT": "5000",
"PD_FEEDBACK_PORT": "5001"
}
}
}
}
Note: If the
uvcommand is not found, runwhich uv(Unix) orGet-Command uv(PowerShell) and use the full path in the"command"field.
Quick Example
Once the MCP server is connected and the Pd patch is listening, you can ask Claude to build patches using natural language:
"Create an osc~ at 440 Hz, connect it to a dac~, and start DSP."
Claude will call the MCP tools to produce something like:
create_object("osc~", ["440"], {"x": 100, "y": 50}) → index 0
create_object("*~", ["0.25"], {"x": 100, "y": 110}) → index 1
create_object("dac~", [], {"x": 100, "y": 170}) → index 2
connect_objects({"id": "osc~_100_50", "port": 0}, {"id": "*~_100_110", "port": 0})
connect_objects({"id": "*~_100_110", "port": 0}, {"id": "dac~_100_170", "port": 0})
connect_objects({"id": "*~_100_110", "port": 0}, {"id": "dac~_100_170", "port": 1})
start_dsp()
This creates the following patch in the workspace subpatch:
osc~ 440
|
*~ 0.25 (volume)
/ \
dac~ L+R
You can then ask Claude to modify the patch, add effects, change frequencies, or build more complex signal chains — all through conversation.
Architecture
Message Flow
- Claude executes MCP tools based on user requests
- MCP Server processes the tool calls
- OSC messages are formatted and sent to Pure Data
- Pure Data executes the commands via the dynamic patch
- Feedback (if any) is returned via OSC callbacks
Troubleshooting
Common Issues
- Object Creation Fails: Ensure Pure Data is running and the patch is open
- Connection Issues: Verify that object IDs match exactly what was returned from
create_object - Port Conflicts: Check if port 5000 is already in use
- Message Format Errors: Ensure message formats match the expected format in the Pure Data patch
- Lost Objects: If object tracking gets confused, try restarting both the MCP server and Pure Data
Known Limitations
- No delete or disconnect support: The Pd patch (
example_patch.pd) only handles/pd/create,/pd/connect,/pd/dsp, and/pd/clear. Objects and connections cannot be individually removed — useclear_workspaceto reset the entire workspace instead. - Signal objects with zero arguments: When signal objects like
+~or*~are created with a0argument via dynamic patching (e.g.,obj 100 100 +~ 0), Pd 0.55+ treats the second inlet as a float inlet instead of a signal inlet, breaking signal connections. The MCP server automatically strips trailing zero arguments from signal objects to avoid this issue.
License
This project is licensed under the MIT License - see the LICENSE file for details.
Acknowledgements
- Pure Data - The open source visual programming language
- Model Context Protocol - The protocol enabling AI tools
- Python-OSC - Python implementation of OSC
Installing Pd
This server has no published package — it is built from source. Open the repository and follow its README.
▸ github.com/nikmaniatis/Pd-MCP-ServerFAQ
Is Pd MCP free?
Yes, Pd MCP is free — one-click install via Unyly at no cost.
Does Pd need an API key?
No, Pd runs without API keys or environment variables.
Is Pd hosted or self-hosted?
Self-hosted: the server runs locally on your machine via the install command above.
How do I install Pd in Claude Desktop, Claude Code or Cursor?
Open Pd on unyly.org, pick your client tab (Claude Desktop, Claude Code, Cursor) and press Install — the config is generated automatically, no JSON editing.
Related MCPs
ARA
Generate images, video and audio from any AI agent — one connector.
by ARAOmni Video
An MCP server that transforms LLM-enabled IDEs into professional video editors by pre-processing footage into text proxies, generating motion graphics via HTML/
by buildwithtazaYouTube
Transcripts, channel stats, search
by YouTubeEverArt
AI image generation using various models.
by modelcontextprotocolCompare Pd with
Not sure what to pick?
Find your stack in 60 seconds
Author?
Embed badge for your README
Browse similar
All media MCPs
