TouchDesigner MCP Server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@TouchDesigner MCP Server write a script to generate a reaction diffusion network"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
TouchDesigner MCP Server
A powerful Model Context Protocol (MCP) server that brings comprehensive TouchDesigner operator documentation, Python API references, and tutorials directly to your AI coding assistant.
I personally use it with VS Code/Codium and the Roo Code Extension alongside my favorite LLMs like Claude Opus and GPT-5. While the server should theoretically work with other applications like Docker or Claude Desktop, I haven't thoroughly tested those integrations yet.
The Story So Far: The catalyst for this project was simple -- A few months ago I decided I wanted an AI tool that could develop visuals in TouchDesigner for me. I've been making some basic things in TD for awhile but it has one of the steepest learning curves of any creative software I've ever used. How nice would it be just to ask an LLM "Please make a 720p, animated in realtime, constantly evolving, reaction diffusion simulation in TouchDesigner." and then just have it spit one out? Super duper nice is the answer in my opinion.
Like every project I undertake where I know very little, I thought it would be pretty straight forward! LOL! When I started I had no idea just how much time and tokens I would end up sinking into this one. And I still have a ways to go before I consider it complete with full integration into TouchDesigner itself and very little margin of error.
However as far as querying information about operators and their parameters goes -- it works pretty darn well! The major challenge I face with getting any LLM to use this tool effectively for developing networks with real complexity is that the models tend to forget to use the MCP server and fall back on their outdated trained knowledge of much older TouchDesigner versions. If anyone has suggestions on how to overcome this particular wrinkle, I'd absolutely love to hear your feedback!
How to Use It (My Current Workflow)
Right now, I'm focused on ensuring the server returns useful and accurate Python API information. Here's my workflow: I describe the network I want to create -- for example, a reaction diffusion visualization -- and then ask the AI to write a Python script that generates the network inside TouchDesigner using the textport. Once that's complete, I get an 'exec' command that lets me copy and paste just one line into the textport, and voila -- a network is generated!
My ultimate dream, however, is full integration directly into TouchDesigner itself. I experimented with a web server and WebSocket approach, which sort of worked, but since the LLMs I was testing weren't generating correct Python code and kept hallucinating information while not utilizing the MCP server to its fullest potential, I've temporarily shelved this part of the project. Once it's working reliably and generating solid visualizations or networks, then full integration will be the final step!
Related MCP server: TD_MCP
Features
661 TouchDesigner Operators - Comprehensive documentation including 102 POP (Point Operator) entries
14 Interactive Tutorials - Comprehensive TouchDesigner learning guides
214 Python API Classes - Full Python scripting documentation with 1,674+ methods
21 MCP Tools - Across five functional groups: operator reference, tutorials, Python API, version system, and experimental content
32 Workflow Patterns - Curated operator chain patterns with 72 common transitions
Current with TouchDesigner 2025 (latest official build 2025.32820, bundled Python 3.11.10) and the 2025.30000 experimental series (Python 3.11.10). POPs (Point Operators) are the flagship new operator family in the 2025 release — the first new operator family in over a decade.
Version System
Version History - All stable TD releases documented from 099 through 2025 (latest official build 2025.32820, Python 3.11.10)
Compatibility Tracking - Per-operator and per-method version compatibility data
Python Timeline - Full Python version history bundled with each TD release
Release Highlights - Key features and breaking changes per major release, rewritten from live Derivative sources with provenance
Experimental Techniques Knowledge Base
7 Technique Categories - GLSL, GPU compute, machine learning, generative systems, audio-visual, networking, Python advanced
2,000+ Lines of Code - Working, paste-ready code snippets for advanced techniques
16 Named GLSL Patterns - Complete shader code for raymarching, reaction-diffusion, feedback, and more
Difficulty Ratings - Each technique rated with minimum TD version requirements
Core Enhancements
Operator Wiring Guides - Exact port-level connection instructions for 20+ common operators
Network Templates - Five ready-to-build network templates with Python generation scripts
Version-Aware Search - Filter operators and Python API by TD release compatibility
Smart Workflow Suggestions - Port wiring, complexity ratings, and node count estimates
Experimental Build Support
Experimental Build Series - The 2025.30000 series (POPs, Python 3.11.10) and the 2021 Vulkan experimental build, sourced from live Derivative release notes
Feature Flag Tables - Opt-in/opt-out flags for each experimental series
Graduation Tracking - Which experimental features graduated into stable releases
Breaking Change Logs - Per-series breaking changes versus the stable baseline
Core Server Features
Smart Search - Direct search with category filtering, parameter search, and version filtering
Zero Configuration - Works immediately after installation
Pure MCP Implementation - Clean stdio-based server, no web server overhead
Local Data Processing - All documentation served from local JSON files, no network requests
Installation
Global Installation (Recommended)
npm install -g @bottobot/td-mcpLocal Installation
npm install @bottobot/td-mcpQuick Start
Using with VS Code/Codium and Claude
Install the MCP server globally:
npm install -g @bottobot/td-mcpConfigure Claude/MCP in VS Code: Add the server to your MCP settings configuration file:
{ "td-mcp": { "command": "npx", "args": ["@bottobot/td-mcp"] } }Start using TouchDesigner tools in your conversations with Claude!
Running Standalone
# If installed globally
td-mcp
# If installed locally
npx @bottobot/td-mcpAvailable MCP Tools (21 Total)
Core Operator Tools
get_operator
Get comprehensive details about a specific TouchDesigner operator including all parameters, tips, and code examples.
Parameter | Type | Required | Description |
| string | Yes | Operator name (e.g., 'Noise CHOP', 'Movie File In TOP') |
| boolean | No | Show code examples and usage |
| boolean | No | Show tips and performance notes |
| string | No | Include compatibility block showing when the operator was added or changed for this TD version |
Example: "Get detailed information about the Noise CHOP operator"
Example: "Get the Engine COMP documentation for TD 2022"search_operators
Search for operators using contextual analysis and relevance ranking.
Parameter | Type | Required | Description |
| string | Yes | Search query |
| string | No | Filter by category (CHOP, DAT, SOP, TOP, MAT, COMP, POP) |
| string | No | Filter by subcategory (e.g., 'Audio', 'Filters') |
| string | No | Search mode: 'fuzzy' (default), 'exact', or 'tag' |
| string | No | Filter to operators compatible with a specific TD version (e.g., '2022') |
| boolean | No | Search within parameter names and descriptions |
| boolean | No | Show detailed results with keywords |
| number | No | Maximum results (default: 10, max: 50) |
Example: "Search for audio processing operators in the CHOP category"
Example: "Find operators added in TD 2022"list_operators
List available TouchDesigner operators with optional category filtering.
Parameter | Type | Required | Description |
| string | No | Filter by operator category |
Example: "List all TOP operators"compare_operators
Compare two operators side by side -- parameters, categories, shared/unique features.
Parameter | Type | Required | Description |
| string | Yes | First operator name |
| string | Yes | Second operator name |
| boolean | No | Include parameter comparison (default: true) |
Example: "Compare Blur TOP with Luma Blur TOP"get_operator_examples
Get Python code examples, expressions, and usage patterns for a specific operator.
Parameter | Type | Required | Description |
| string | Yes | Operator name |
| string | No | Type: 'all', 'python', 'expressions', or 'usage' |
Example: "Get Python examples for Movie File In TOP"suggest_workflow
Get workflow suggestions for what operators commonly follow the current operator.
Parameter | Type | Required | Description |
| string | Yes | Current operator name |
Example: "What operators typically follow a Movie File In TOP?"Tutorial Tools
get_tutorial
Access detailed TouchDesigner tutorial content with full sections, code, and links.
Parameter | Type | Required | Description |
| string | Yes | Tutorial name |
| boolean | No | Include full content sections |
| boolean | No | Include table of contents |
| boolean | No | Include related links |
Example: "Get the 'Write a GLSL TOP' tutorial"list_tutorials
List all available TouchDesigner tutorials with optional filtering.
Parameter | Type | Required | Description |
| string | No | Search term to filter tutorials |
| number | No | Maximum results |
| boolean | No | Show tutorial summaries and keywords |
Example: "List all available tutorials"search_tutorials
Search through tutorial content by keyword, topic, or content.
Parameter | Type | Required | Description |
| string | Yes | Search query |
| boolean | No | Search within section content (default: true) |
| number | No | Maximum results (default: 10) |
Example: "Search tutorials for GLSL shader examples"Python API Tools
get_python_api
Get documentation for a TouchDesigner Python class including members and methods.
Parameter | Type | Required | Description |
| string | Yes | Python class name (e.g., 'CHOP', 'Channel', 'App') |
| boolean | No | Show class members/properties |
| boolean | No | Show class methods |
| boolean | No | Show inherited members and methods |
| string | No | Annotate each method/member with its introduction version and exclude API added after this TD version |
Example: "Get Python documentation for the CHOP class"
Example: "Get Python API for the App class as of TD 2021"search_python_api
Search across TouchDesigner Python classes, methods, and members.
Parameter | Type | Required | Description |
| string | Yes | Search query |
| string | No | Where to search: 'all', 'classes', 'methods', 'members' |
| string | No | Filter by category |
| string | No | Filter classes, methods, and members to those available in the specified TD version |
| number | No | Maximum results |
Example: "Search Python API for audio methods"
Example: "Find Python API methods available in TD 2020"list_python_classes
List all available Python API classes grouped by category.
Parameter | Type | Required | Description |
| string | No | Filter by category (e.g., 'Operator', 'General') |
| string | No | Search term to filter classes |
| boolean | No | Show member/method counts |
Example: "List all Python API classes in the Operator category"Version System Tools
get_version_info
Get detailed information about a specific TouchDesigner stable release: which Python version it bundles, new operators introduced, key features, Python API additions, and breaking changes.
Parameter | Type | Required | Description |
| string | Yes | TD version string (e.g., '2025', '2022', '2019', '099') |
Example: "What Python version does TouchDesigner 2025 use?"
Example: "What operators were added in TouchDesigner 2025?"list_versions
List all supported TD versions (099, 2019, 2020, 2021, 2022, 2023, 2025) with a quick-reference table showing the bundled Python version and support status for each release, plus the full Python version timeline.
Parameter | Type | Required | Description |
(none required) | — | — | Returns all versions with highlights |
Example: "List all supported TouchDesigner versions"
Example: "Show the Python version timeline for TouchDesigner"Experimental Techniques Tools
get_experimental_techniques
Browse a curated library of advanced TouchDesigner techniques by category. Returns descriptions, difficulty ratings, version requirements, operator chains, uniform tables, and full working code snippets.
Parameter | Type | Required | Description |
| string | Yes | Technique category. Aliases accepted: 'glsl', 'gpu-compute', 'machine-learning', 'generative-systems', 'audio-visual', 'networking', 'python-advanced' |
| string | No | ID of a specific technique within the category |
Example: "Show me GLSL raymarching techniques in TouchDesigner"
Example: "Get GPU compute techniques using numpy"Available Categories:
Category | Alias(es) | Techniques |
glsl | shader, raymarching, sdf | Raymarching, reaction-diffusion, feedback, procedural noise |
gpu-compute | gpu, cuda | Script TOP numpy, CUDA, Shared Memory, GPU instancing |
machine-learning | ml, ai | Engine COMP, ONNX, Stable Diffusion, MediaPipe, Body Track |
generative-systems | generative, lsystem | L-systems, Game of Life, strange attractors, boids |
audio-visual | audio, fft | FFT geometry, beat detection, granular synthesis, MIDI |
networking | network, osc, ndi | OSC, WebSocket, NDI, TDAbleton, multi-machine |
python-advanced | python, numpy, opencv | asyncio, tdu.Dependency, threading, numpy, OpenCV |
search_experimental
Full-text search across all 7 experimental technique categories. Results ranked by weighted field scoring (name, tags, description, notes, code).
Parameter | Type | Required | Description |
| string | Yes | Search query |
| string | No | Restrict search to one category |
| boolean | No | Include code snippets in results |
| number | No | Maximum results (default: 10, max: 30) |
Example: "Search experimental techniques for reaction diffusion"
Example: "Find GPU instancing examples in the experimental library"get_glsl_pattern
Retrieve a specific named GLSL pattern with complete, paste-ready shader code. Covers 16
named patterns across raymarching, reaction-diffusion, feedback, procedural noise, cellular
automata, and GPU particle simulation. Also provides three reusable GLSL utility libraries.
Use pattern: "list" to enumerate all available patterns.
Parameter | Type | Required | Description |
| string | Yes | Pattern name or 'list' to see all available patterns |
| boolean | No | Include GLSL utility library code (default: false) |
Example: "Get the raymarching GLSL pattern"
Example: "List all available GLSL patterns"
Example: "Get the reaction diffusion shader pattern"Core Enhancement Tools
get_operator_connections
Get a wiring guide for a specific operator: what operators typically connect upstream as inputs and downstream as outputs, with exact port numbers, rationale, and workflow pattern names. Covers 20+ common operators across all families.
Parameter | Type | Required | Description |
| string | Yes | Operator name (with or without family suffix; case-insensitive) |
Example: "What connects to and from a Render TOP?"
Example: "Show me the typical connections for a Noise CHOP"get_network_template
Return a complete, ready-to-use network template for a common TouchDesigner use case.
Each template includes an operator list, a port-level connection table, parameter settings,
and a ready-to-paste Python script that builds the network. Use template: "list" to see
all available templates.
Parameter | Type | Required | Description |
| string | Yes | Template name or 'list' to enumerate all templates |
Available Templates:
Template | Description |
| File-based video playback with level and output controls |
| Noise-driven generative visual network |
| Audio analysis feeding visual parameters |
| Table DAT driven chart and display network |
| Multi-layer compositing setup for live use |
Example: "Give me a network template for audio-reactive visuals"
Example: "List all available network templates"Experimental Build Tools
TouchDesigner ships two parallel release tracks: stable releases (the official line went 099 → 2019 → 2020 → 2021 → 2022 → 2023 → 2025; there was never an official TD 2024) and experimental/beta build series with unreleased features. These tools give you full access to the experimental track.
get_experimental_build
Get detailed information about a specific experimental TD build series or the latest experimental series.
Parameter | Type | Required | Description |
| string | No | Experimental series ID (e.g. '2025.10000'). Omit for the latest experimental series. |
| boolean | No | Include new features list (default: true) |
| boolean | No | Include breaking changes vs stable (default: true) |
| boolean | No | Include Python API additions (default: true) |
| boolean | No | Include experimental operators (default: true) |
Example: "What new features are in the latest experimental TouchDesigner build?"
Example: "Get breaking changes for experimental series 2025.30000"list_experimental_builds
List recent experimental TD build series grouped by feature area (rendering, Python API, operators, UI, networking).
Parameter | Type | Required | Description |
| string | No | Filter by area: 'rendering', 'Python API', 'operators', 'UI', 'networking' |
| string | No | 'experimental' (active) or 'graduated' (became stable) |
| boolean | No | Include feature flag tables (default: false) |
| boolean | No | Include experimental operator lists (default: true) |
| boolean | No | Include breaking change summaries (default: false) |
Example: "List experimental TD builds with rendering changes"
Example: "Which experimental build series introduced POPs?"Tracked Experimental Series:
Series ID | Year | Status | Headline Feature |
2025.30000 | 2025 | Graduated (TD 2025) | POPs (Point Operators) flagship family, Python 3.11.10 |
2021.30000 | 2021 | Graduated (TD 2022) | Vulkan becomes sole graphics API (OpenGL retired), Python 3.9.5 |
Both series are sourced from the live Derivative release notes (provenance recorded in
wiki/data/versions/experimental-builds.json). The 2025.30000 series led into the official
2025 release; the 2021 experimental build first moved TouchDesigner to Vulkan, which became
the default in the 2022 release.
Operator Categories
Category | Count | Description |
CHOP | 170 | Channel Operators - Audio, control signals, and data streams |
TOP | 147 | Texture Operators - 2D image and video processing |
SOP | 113 | Surface Operators - 3D geometry creation and manipulation |
POP | 102 | Point Operators - Point/particle geometry (new in TD 2025) |
DAT | 75 | Data Operators - Text, tables, and data handling |
COMP | 41 | Component Operators - UI elements and containers |
MAT | 13 | Material Operators - 3D rendering materials and shaders |
Total | 661 | All seven operator families |
Available Tutorials
Core Tutorials
Anatomy of a CHOP - Understanding channel operator internals
Build a List COMP - Create dynamic UI lists
Introduction to Python Tutorial - Essential guide for TouchDesigner Python scripting
Advanced Development
Write a GLSL TOP - Create custom GPU-accelerated image effects
Write a GLSL Material - Custom GLSL materials for rendering
Write a C++ CHOP - Develop native audio/data processing operators
Write a C++ TOP - Build custom texture operators in C++
Write a C++ Plugin - Advanced C++ integration for TouchDesigner plugins
Write a CUDA DLL - GPU programming with CUDA for TouchDesigner
Inter-Process Communication
Write a Shared Memory CHOP - Inter-process data communication
Write a Shared Memory TOP - Share textures between processes
Video & Integration
Video Streaming User Guide - Complete guide for video streaming workflows
TouchDesigner Video Server Specification Guide - Professional video server setup
TDBitwig User Guide - Integration with Bitwig Studio DAW
Python API Documentation
The server includes documentation for 214 Python API classes with 1,674+ methods covering:
Core operator classes (CHOP, TOP, SOP, DAT, MAT, COMP)
Utility classes (Channel, Cell, Page, etc.)
System classes (App, Project, Monitor, etc.)
UI classes (Panel, Widget, etc.)
Advanced features (WebRTC, NDI, MIDI, OSC, etc.)
Architecture
The TD-MCP server is built with:
Pure MCP Implementation - Clean stdio-based server following MCP standards
Direct Search Algorithm - Fast, reliable search without external index dependencies
OperatorDataManager - Centralized data management with 661 operators loaded into memory
Local Data Processing - All operator data is processed and served locally
Modular Tool System - Each of the 21 MCP tools is independently maintained
Dual Release Track Support - Stable releases (099–2025) and experimental build series both fully documented
Project Structure
td-mcp/
├── index.js # Main MCP server entry point (21 tools)
├── tools/ # MCP tool implementations (21 tools)
│ ├── get_operator.js # Full operator documentation
│ ├── search_operators.js # Operator search with ranking
│ ├── suggest_workflow.js # Workflow chain suggestions
│ ├── list_operators.js # List/filter operators
│ ├── get_tutorial.js # Tutorial content access
│ ├── list_tutorials.js # Tutorial listing
│ ├── search_tutorials.js # Tutorial content search
│ ├── get_python_api.js # Python class documentation
│ ├── search_python_api.js # Python API search
│ ├── list_python_classes.js # Python class browsing
│ ├── get_operator_examples.js # Code examples per operator
│ ├── compare_operators.js # Side-by-side comparison
│ ├── get_version_info.js # Stable version details
│ ├── list_versions.js # All stable versions
│ ├── get_experimental_techniques.js # Advanced technique library
│ ├── search_experimental.js # Search techniques
│ ├── get_glsl_pattern.js # Named GLSL shader patterns
│ ├── get_operator_connections.js # Operator wiring guide
│ ├── get_network_template.js # Full network templates
│ ├── get_experimental_build.js # Experimental build series details
│ └── list_experimental_builds.js # List experimental series by area
├── wiki/ # Documentation system
│ ├── data/
│ │ ├── processed/ # 661 operator JSON files
│ │ ├── tutorials/ # 14 tutorial JSON files
│ │ ├── python-api/ # 214 Python class JSON files
│ │ ├── experimental/ # 7 advanced technique JSON files
│ │ └── versions/ # Version compatibility data
│ │ ├── version-manifest.json
│ │ ├── operator-compatibility.json
│ │ ├── python-api-compatibility.json
│ │ ├── release-highlights.json
│ │ └── experimental-builds.json # Experimental track data
│ ├── utils/
│ │ └── version-filter.js # Version utilities + experimental support
│ ├── operator-data-manager.js # Core documentation engine
│ └── operator-data-python-api.js # Python API data manager
├── data/
│ └── patterns.json # 32 workflow patterns + transitions
├── scripts/ # Data maintenance scripts
│ ├── clean-operator-data.js # Clean parameter descriptions
│ └── enrich-top-operators.js # Add tips/examples to operators
└── package.jsonTroubleshooting
Server won't start
Ensure Node.js 20.0+ is installed:
node --versionTry reinstalling:
npm install -g @bottobot/td-mcpCheck for port conflicts if running other MCP servers
Search returns no results
Try broader search terms
Remove category filters
Enable
parameter_searchto search within parameter namesCheck spelling of operator names
Operator not found
Use the full operator name including family:
Noise CHOPnot justNoiseUse
search_operatorsfirst to find the exact nameUse
list_operatorswith a category filter to browse available operators
LLM not using the MCP tools
Explicitly mention TouchDesigner in your prompt
Ask the LLM to "use the td-mcp tools" to look up operator information
Remind the LLM to check parameters with
get_operatorbefore generating code
Requirements
Node.js 20.0 or higher
npm package manager
VS Code/Codium with MCP-compatible extension (e.g., Claude Dev, Roo Code)
Development
To contribute or modify the server:
# Clone the repository
git clone https://github.com/bottobot/touchdesigner-mcp-server.git
# Install dependencies
cd touchdesigner-mcp-server
npm install
# Run the server locally
node index.js
# Validate syntax
node --check index.js
# Clean operator data (removes HTML dumps from parameter descriptions)
node scripts/clean-operator-data.js
# Enrich top operators with tips and examples
node scripts/enrich-top-operators.jsAdding a New Tool
Create a new file in
tools/following the existing pattern (exportschemaandhandler)Import and register the tool in
index.jsUpdate this README with the tool documentation
Adding Operator Data
Operator data lives in wiki/data/processed/ as JSON files. Each file follows the schema with fields: id, name, displayName, category, subcategory, description, parameters, tips, warnings, pythonExamples, codeExamples, and version.
Support
Issues: GitHub Issues
Documentation: TouchDesigner Official Docs
NPM Package: @bottobot/td-mcp
License
MIT License - See LICENSE file for details.
Acknowledgments
TouchDesigner by Derivative
Model Context Protocol by Anthropic
Built for the TouchDesigner community
Current Version: 3.0.0 Operators: 661 (CHOP 170, TOP 147, SOP 113, POP 102, DAT 75, COMP 41, MAT 13) Tutorials: 14 Python API Classes: 214 MCP Tools: 21 Workflow Patterns: 32 TouchDesigner Currency: 2025 official (build 2025.32820, Python 3.11.10) + 2025.30000 experimental Last Updated: June 2026
Available Tools
37 toolscompare_operatorsCompare TouchDesigner OperatorsB
Compare two TouchDesigner operators side by side - parameters, categories, and use cases
Inputs: operator_a (required), operator_b (required), compare_parameters.
get_experimental_buildGet TouchDesigner Experimental Build InfoA
Returns information about a TouchDesigner experimental/beta build series. Experimental builds are pre-release tracks that contain new features not yet available in stable releases. You can request a specific series ID (e.g. '2025.30000') or omit the parameter to get the latest experimental series. Response includes: new features, experimental operators, breaking changes versus the stable baseline, and Python API additions.
Inputs: series_id, show_features, show_operators, show_python_api, +1 more.
get_experimental_techniquesGet Experimental TechniquesA
Browse TouchDesigner experimental techniques by category. Categories: glsl, gpu-compute, machine-learning, generative-systems, audio-visual, networking, python-advanced. Returns technique descriptions, code snippets, operator chains, and setup notes.
Inputs: category (required), show_code, show_setup, technique_id.
get_glsl_patternGet GLSL PatternA
Retrieve specific GLSL shader patterns or GPU technique code snippets with full code ready to paste into a TouchDesigner GLSL TOP. Also provides common GLSL utility libraries. Use 'list' to see all available patterns.
Inputs: pattern (required), include_utilities.
get_network_templateGet TouchDesigner Network TemplateA
Returns a full TouchDesigner network template for a common use case. Each template includes the operator list, wire connections with port numbers, parameter settings, ready-to-paste Python scripts, and setup tips. Available templates: video-player, generative-art, audio-reactive, data-visualization, live-performance.
Inputs: template (required), show_python.
get_operatorGet TouchDesigner OperatorB
Get comprehensive details about a specific TouchDesigner operator. Optionally pass a TouchDesigner version to see compatibility notes for that release.
Inputs: name (required), version, show_tips, show_examples.
get_operator_connectionsGet Operator ConnectionsA
Returns the typical upstream inputs, downstream outputs, and common operator chains for a named TouchDesigner operator. Includes connection port instructions (e.g., 'A output 0 -> B input 0') so you know exactly how to wire nodes together.
Inputs: operator_name (required), direction.
get_operator_examplesGet TouchDesigner Operator ExamplesB
Get Python code examples, expressions, and usage patterns for a specific TouchDesigner operator
Inputs: operator (required), example_type.
get_python_apiGet Python API DocumentationA
Get documentation for a TouchDesigner Python class. Optionally filter to show only API available in a specific TD version.
Inputs: class_name (required), version, show_members, show_methods, +1 more.
get_tutorialGet TouchDesigner TutorialB
Get detailed content from a TouchDesigner tutorial
Inputs: name (required), include_toc, include_links, include_content.
get_version_infoGet TouchDesigner Version InfoA
Get detailed information about a specific TouchDesigner version including Python version, new operators, key features, and API changes.
Inputs: version (required), show_new_operators, show_breaking_changes, show_python_highlights.
list_experimental_buildsList TouchDesigner Experimental BuildsA
Lists recent TouchDesigner experimental/beta build series grouped by feature area. Feature areas: rendering, Python API, operators, UI, networking. Each entry shows the series ID, build range, stability status, and headline features. Use get_experimental_build with a series_id for full details on any individual series.
Inputs: feature_area, show_operators, stability_status, show_feature_flags, +1 more.
list_operatorsList TouchDesigner OperatorsC
List available TouchDesigner operators with contextual grouping
Inputs: category.
list_python_classesList TouchDesigner Python ClassesA
List all available TouchDesigner Python API classes, optionally filtered by category
Inputs: search, category, show_details.
list_tutorialsList TouchDesigner TutorialsB
List available TouchDesigner tutorials with summaries
Inputs: limit, search, show_details.
list_versionsList TouchDesigner VersionsB
List all supported TouchDesigner versions with Python versions, support status, and key feature highlights for each release.
Inputs: filter_status, show_highlights, show_python_info.
search_experimentalSearch Experimental TechniquesB
Search across all TouchDesigner experimental technique categories. Finds techniques by keyword, operator name, tag, or code pattern across GLSL, GPU compute, machine learning, generative systems, audio-visual, networking, and advanced Python.
Inputs: query (required), limit, show_code, category_filter.
search_operatorsSearch TouchDesigner OperatorsA
Search for operators using advanced contextual analysis and ranking. Supports multi-term queries, category/subcategory filtering, parameter search, typo tolerance, search type selection (exact/fuzzy/tag), and optional TouchDesigner version compatibility filtering.
Inputs: query (required), type, limit, version, +4 more.
search_python_apiSearch Python APIA
Search across TouchDesigner Python classes, methods, and members. Optionally filter by TouchDesigner version to see only API available in that release.
Inputs: query (required), limit, version, category, +1 more.
search_tutorialsSearch TouchDesigner TutorialsC
Search through TouchDesigner tutorials by keyword, topic, or content
Inputs: query (required), limit, search_content.
suggest_workflowSuggest Next TouchDesigner OperatorsA
Get workflow suggestions for what operators commonly follow the current operator. Returns connection port instructions (e.g., 'output 0 -> input 0'), complexity rating (simple/medium/complex), estimated node count for the full workflow, and minimum TD version required.
Inputs: current_operator (required).
td_build_glslTD Build GLSL (live)A
COMPILE a GLSL shader technique from the experimental GLSL knowledge base into a running TouchDesigner instance via the td_mcp bridge. For a SUPPORTED technique (its setup block has machine-readable operators + uniforms) it creates a Text DAT with the shader, a GLSL TOP pointed at that DAT, declares the uniforms, creates supporting operators and lays out. For an UNSUPPORTED technique (multi-pass/feedback systems needing hand wiring) it returns the shader code plus an explanation and does NOT fabricate connections. Builds into the sandbox COMP by default (/td_mcp/sandbox). Use technique:'list' to enumerate. Requires the bridge (see td_status).
Inputs: technique (required), parent.
td_build_patternTD Build Pattern (live)A
COMPILE a linear workflow pattern from data/patterns.json into a running TouchDesigner instance via the td_mcp bridge. It creates each operator in the pattern's workflow order and wires them as a STRAIGHT CHAIN (each op's output 0 into the next op's input 0), then lays out the network. LIMITATION: this only builds LINEAR SINGLE-INPUT chains — patterns.json carries no port/branch data, so multi-input nodes (Composite, Switch, Merge, Render, etc.) are NOT branch-wired; use td_build_template for branched networks. Builds into the sandbox COMP by default (/td_mcp/sandbox). STRICT: hard-errors if any operator name has no createType in the operator map. Requires the bridge (see td_status).
Inputs: pattern (required), parent.
td_build_templateTD Build Template (live)A
COMPILE a named TouchDesigner network template into a running TouchDesigner instance via the td_mcp bridge: it creates every operator, sets the template parameters, wires the connections (port-level), and lays out the network. Templates: video-player, generative-art, audio-reactive, data-visualization, live-performance (aliases like 'feedback loop', 'vj', 'data viz' accepted). Builds into the sandbox COMP by default (/td_mcp/sandbox). STRICT: if any operator type or parameter name cannot be resolved against the operator map, it fails and lists what is missing instead of guessing. Requires the bridge to be running (see td_status).
Inputs: template (required), parent.
td_clearTouchDesigner Clear NetworkA
Delete every operator inside the live-control sandbox COMP. Only clears operators within the sandbox (default /td_mcp/sandbox) — it will not touch the rest of your TouchDesigner project. Requires the TouchDesigner bridge to be running (see td_status).
Inputs: parent.
td_connectTD Connect (live)A
Connect one operator's output to another operator's input in a running TouchDesigner instance (live control via the td_mcp bridge). Wraps the documented Connector.connect() call (op(to).inputConnectors[toIn].connect(op(from).outputConnectors[fromOut])).
Inputs: to_path (required), from_path (required), to_input, from_output.
td_create_operatorTD Create Operator (live)A
Create a new operator inside a parent COMP in a running TouchDesigner instance (live control via the td_mcp bridge). Wraps the documented COMP.create(opType, name) call. op_type may be a create token like 'noiseTOP' or a display name like 'Noise TOP' (resolved via the operator map). Defaults to creating inside the '/td_mcp/sandbox' COMP.
Inputs: op_type (required), name, parent.
td_deleteTD Delete (live)A
Delete (destroy) an operator in a running TouchDesigner instance (live control via the td_mcp bridge). Wraps the documented op.destroy() call. This permanently removes the operator and any of its children. Target a path inside '/td_mcp/sandbox' to stay within the safe work area.
Inputs: path (required).
td_get_errorsTD Get Errors (live)A
Read cook errors, warnings, and script errors from an operator in a running TouchDesigner instance (live control via the td_mcp bridge). Returns the documented op.errors()/op.warnings()/op.scriptErrors() output. Use this to diagnose why a node is not cooking after a build/edit.
Inputs: path (required), recurse.
td_layoutTouchDesigner Layout NetworkA
Auto-arrange the operators inside a TouchDesigner network so the node graph is tidy and readable. Operates on the sandbox COMP by default (/td_mcp/sandbox). Requires the TouchDesigner bridge to be running (see td_status).
Inputs: parent.
td_list_networkTouchDesigner List NetworkA
List the operators inside a TouchDesigner network (a parent COMP). Returns each child's path, type, node position, input/output counts and error counts. Defaults to the live-control sandbox COMP (/td_mcp/sandbox). Requires the TouchDesigner bridge to be running (see td_status).
Inputs: parent, max_depth.
td_renderTD Render (live)A
Render the current frame of a TOP in a running TouchDesigner instance and return the image so the assistant can see the result (live control via the td_mcp bridge). Uses the documented TOP.saveByteArray() call. Use this after building/editing a network to visually verify the output.
Inputs: path (required), filetype.
td_run_pythonTD Run Python (live, UNSAFE)A
Run ARBITRARY Python code inside a running TouchDesigner instance via the td_mcp bridge. WARNING: this is arbitrary code execution with full access to the TouchDesigner project and the host Python environment. It is DISABLED BY DEFAULT and only succeeds when the bridge's "Allow Python" flag is explicitly turned on; otherwise the bridge rejects the request. Prefer the structured live-control tools (td_build_template, td_set_parameter, td_connect, td_list_network, ...) for normal operations and only use this for ad-hoc Python that those cannot express. Returns stdout / the returned value plus any errors, warnings and scriptErrors from the bridge.
Inputs: code (required).
td_sampleTD Sample (live)A
Sample a single value from an operator in a running TouchDesigner instance (live control via the td_mcp bridge). kind='chop_channel' reads a CHOP channel via CHOP.chan(name); kind='top_pixel' reads a pixel via TOP.sample(u, v). Use this to verify concrete output values without rendering a full frame.
Inputs: kind (required), path (required), u, v, +1 more.
td_set_parameterTD Set Parameter (live)A
Set a parameter on an operator in a running TouchDesigner instance (live control via the td_mcp bridge). Drives the documented op.par.NAME.val (value), op.par.NAME.expr (expr) and op.par.NAME.pulse() members. parameter may be a documented label (e.g. 'Period') or an exact parameter name (e.g. 'period'); labels are resolved via the operator map. Provide exactly one of value, expr, or pulse.
Inputs: path (required), parameter (required), expr, pulse, +1 more.
td_set_resolutionTD Set Resolution (live)A
Set the output resolution (width x height in pixels) of a TOP in a running TouchDesigner instance (live control via the td_mcp bridge). Targets a TOP's documented resolution parameters.
Inputs: path (required), width (required), height (required).
td_statusTouchDesigner Live StatusA
Check whether the live-control TouchDesigner bridge is reachable and the configured token is accepted. Returns a short status report, or an install/setup hint when the bridge cannot be reached. Reads connection settings from TD_MCP_HOST, TD_MCP_PORT and TD_MCP_TOKEN.
Inputs: no named parameters.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 37 tools
Most tools have distinct purposes, especially between query tools (get/list/search) and live-control tools (td_*). Some overlap exists between get_operator, search_operators, and list_operators, but descriptions clarify scope.
Naming conventions are mixed: query tools use verbs like get, list, search, suggest, while live-control tools consistently use td_ prefix. The inconsistency is noticeable but still readable.
37 tools is on the high side for a single MCP server. The domain is complex and the count may be justified, but it could be split into reference and live-control subsets for better focus.
The tool set covers both informational needs (operators, tutorials, versions, experimental features) and live network building (create, connect, set parameters, render, sample, errors). There are no obvious gaps for the stated purpose.
Maintenance
Related MCP Connectors
Search and query nTop's knowledge base and engineering guides from AI applications.
Provide your AI coding tools with token-efficient access to up-to-date technical documentation for…
Bring your Knacco library into your AI tool: browse, search, read guides and download files.
Versioned documentation registry and semantic search for AI tools and coding assistants.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables natural language control of TouchDesigner via AI, allowing operator creation, parameter setting, Python execution, and node graph building.7MIT
- AlicenseNot gradedqualityDmaintenanceA local-first MCP server for TouchDesigner that provides offline documentation retrieval and live control of TouchDesigner sessions, enabling AI-assisted network building and parameter management.2MIT
- AlicenseBqualityDmaintenanceEnables AI assistants to build and control TouchDesigner networks by describing visuals in natural language, with automatic node creation, verification, and preview.100MIT
- AlicenseAqualityBmaintenanceEnables AI agents to index TouchDesigner installations, control live instances with undoable batches and health checks, and inspect saved .toe projects offline.46MIT