SuperCollider MCP Server
A Model Context Protocol (MCP) server for controlling SuperCollider audio synthesis. This server enables AI assistants like Claude to generate and control real-time audio synthesis through natural language descriptions.
Features
Natural Language Sound Synthesis: Describe sounds in plain English and have them synthesized in real-time
10 Built-in Synth Types: sine, pluck, bell, bass, pad, kick, snare, hihat, atmosphere, sweep
Server Lifecycle Management: Boot, quit, and monitor SuperCollider server status
Pattern Sequencing: Create rhythmic patterns and melodic sequences
Audio Recording: Record synthesis output to files
Raw Code Execution: Execute custom SuperCollider code for advanced control
TypeScript & Type-Safe: Fully typed with Zod schema validation
Prerequisites
Node.js 18+
SuperCollider installed locally (download here)
Installing SuperCollider
macOS:
Download from https://supercollider.github.io/downloads
Drag
SuperCollider.app
to/Applications/
The MCP server will auto-detect it
Linux (Ubuntu/Debian):
Linux (Arch):
Windows:
Download installer from https://supercollider.github.io/downloads
Run the installer
The MCP server will auto-detect it
Custom Installation Path
If SuperCollider is installed in a non-standard location, set the environment variable:
Installation
Usage
With Claude Desktop
Add to your Claude Desktop config file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Restart Claude Desktop after updating the config.
With MCP Inspector (for testing)
Standalone Development
Available Tools
Setup & Diagnostics
sc_health_check
Check if SuperCollider is installed and configured correctly. Run this first if you have any issues.
Returns:
✅ Success: SuperCollider path and confirmation
❌ Error: Installation instructions for your platform
Core Server Management
sc_boot
Boot the SuperCollider audio server. Must be called before any sound synthesis.
sc_quit
Quit the SuperCollider server and clean up resources.
sc_status
Get current server status including CPU usage and running synths.
Sound Synthesis
sc_play_synth
(Recommended for AI)
Play a synthesized sound based on a natural language description. This is the primary tool for sound generation.
Examples:
"play a bell sound at C4"
"play a low bass tone for 2 seconds"
"play a bright plucked string"
"play a soft atmospheric pad"
"play a short snare"
"play a sweep from 100hz to 2000hz over 3 seconds"
The parser understands:
Note names: C4, D#5, Bb3, etc.
Frequencies: 440hz, 880hz, etc.
Descriptive terms: high, low, bright, deep, soft, loud, short, long
Synth types: bell, bass, pad, pluck, kick, snare, hihat, sweep, atmosphere
Spatial terms: left, right (for panning)
sc_play_synth_advanced
Play a specific synth with explicit parameters for precise control.
Available synths: sine, pluck, bell, bass, pad, kick, snare, hihat, atmosphere, sweep
Parameters:
freq
: Frequency in Hz (default: 440)amp
: Amplitude 0-1 (default: 0.3)duration
: Duration in seconds (default: 1)pan
: Pan position -1 (left) to 1 (right) (default: 0)decay
: Decay time for pluck synth (default: 2)cutoff
: Filter cutoff frequency for bass/padstartFreq
,endFreq
: For sweep synth
sc_play_pattern
Play a rhythmic pattern or sequence of notes.
sc_stop_all
Stop all currently playing synths immediately.
Advanced Control
sc_execute
Execute raw SuperCollider code for custom synthesis and advanced control.
Recording
sc_record_start
Start recording audio output to a file.
Files are saved to the recordings/
directory.
sc_record_stop
Stop the current recording.
Example Conversation Flow
User: "I want to hear a dreamy bell sound"
Claude:
The server will synthesize a bell-like tone using FM synthesis.
User: "Now play a bass line"
Claude:
Synth Descriptions
sine
Pure sine wave tone. Clean and simple.
pluck
Plucked string simulation using the Karplus-Strong algorithm. Good for guitar-like sounds.
bell
Bell-like FM synthesis. Bright and metallic.
bass
Filtered sawtooth wave. Deep bass tones.
pad
Warm, detuned sawtooth pad. Good for ambient backgrounds.
kick
Kick drum with frequency envelope.
snare
Snare drum with noise and tone components.
hihat
Hi-hat cymbal using filtered noise.
atmosphere
Atmospheric noise with slow filter modulation. Good for ambient textures.
sweep
Frequency sweep/riser. Goes from startFreq to endFreq over duration.
Architecture
Key Components:
SuperColliderServer: Manages sclang process lifecycle and code execution
Synth Library: Pre-defined SynthDefs for common sound types
NLP Parser: Converts natural language to synth parameters
MCP Tools: Exposed functions for AI interaction
Integration with MXF
This server is designed to work with the Model Exchange Framework (MXF) for multi-agent music composition. Future enhancements will include:
Event-driven synthesis triggered by MXF events
Shared memory for musical context (key, tempo, style)
Coordination with MIDI MCP server for external control
Agent collaboration on compositional tasks
Development
Build
Watch mode (for development)
Type checking
Troubleshooting
"sclang not found"
Make sure SuperCollider is installed and sclang
is in your PATH:
On macOS, you may need to add to your PATH:
Server won't boot
Check that no other SuperCollider instances are running
Verify SuperCollider works standalone:
sclang
Check server output in the MCP logs
No sound output
Verify your system audio is working
Check SuperCollider audio device settings
Try
Server.default.options.device = "your-device-name";
viasc_execute
Future Enhancements
Real-time parameter modulation
MIDI input/output integration
Sample playback and manipulation
Effects processing (reverb, delay, filters)
Visual waveform/spectrum display
Multi-server support (supernova)
Preset management
MXF event bus integration
Collaborative composition with multiple agents
License
MIT
Contributing
Contributions welcome! Please open an issue or PR.
local-only server
The server can only run on the client's local machine because it depends on local resources.
Tools
Enables AI assistants to generate and control real-time audio synthesis through natural language descriptions using SuperCollider. Features 10 built-in synth types, pattern sequencing, audio recording, and server lifecycle management for creating sounds from simple English descriptions.