touchdesigner

TouchDesigner MCP Server exposes a self-contained MCP interface that serves comprehensive documentation for TouchDesigner operators, Python API references, tutorials, and workflow guidance directly to your AI tooling. It operates as a pure MCP server with local data, meaning all operator docs, Python API details, and tutorial content are served from local JSON data files, avoiding external network requests. This makes it well-suited for integration with code assistants like VS Code/Codium and LLMs such as Claude Opus or GPT-5, enabling you to query operator details, search by category or version, and retrieve ready-to-use code snippets for building TouchDesigner networks. The server is designed to be zero-configuration and quick to stand up after installation, so you can start exploring operator references and Python scripting capabilities with minimal setup.

Prerequisites:

• Node.js and npm installed on your system.
• Basic familiarity with running command-line tools.

Installation steps:

1. Install the MCP server globally (recommended):

   npm install -g @bottobot/td-mcp
   

2. Verify installation by running the server directly (standalone):

   td-mcp
   

   or, if you installed locally:

   npx @bottobot/td-mcp
   

3. If you prefer to integrate with an editor or an external tool, reference the MCP server in your configuration as shown below in the mcp_config snippet.

4. Optional: keep your local data up to date by periodically refreshing the server content if you’re managing a custom data set (not required for standard usage).

Tips and considerations:

• The server focuses on TouchDesigner operator documentation, Python API references, and tutorials. It includes core operator references (e.g., get_operator), search capabilities (e.g., search_operators), and listing utilities (e.g., list_operators). The content is served locally, which helps reduce model hallucinations related to outdated online references.
• If you’re using this with an editor or LLM, you can configure the MCP endpoint in your settings to point to the npx-based startup command shown in mcp_config. This lets the AI tool query operator details, search by TD version, and fetch Python API usage blocks.
• Ensure your TouchDesigner version compatibility is considered when querying for version-specific details. The server provides version-aware search and per-operator/version compatibility data.
• The MCP server is designed to be lightweight and fast with zero configuration; however, you can extend or customize the local JSON data if you want to augment operator docs or add new tutorials and patterns.