RhinoMCP

# RhinoMCP Architecture This document outlines the architecture and component interactions of the RhinoMCP system, which connects Rhino3D to Claude AI via the Model Context Protocol. ## System Overview RhinoMCP consists of three main components that work together to enable AI-assisted 3D modeling: 1. **Rhino Plugin**: A socket server running inside Rhino's Python environment 2. **Rhino Client**: A Python client that communicates with the Rhino plugin 3. **MCP Server**: A WebSocket server implementing the Model Context Protocol These components interact in the following way: ``` Claude AI (Desktop/Windsurf) <--> MCP Server <--> Rhino Client <--> Rhino Plugin <--> Rhino3D ``` ## Component Architecture ### 1. Rhino Plugin (`src/rhino_plugin/`) The Rhino Plugin is a socket server that runs inside Rhino's Python editor environment. It serves as the interface to Rhino's functionality. #### Key Files: - `rhino_server.py`: Socket server implementation that listens for commands and executes them in Rhino #### Responsibilities: - Accept socket connections from external Python processes - Receive commands in JSON format - Execute commands in Rhino's context - Return results or errors in JSON format - Manage error handling and recovery ### 2. Rhino Client (`src/rhino_mcp/rhino_client.py`) The Rhino Client is a Python module that communicates with the Rhino Plugin via a socket connection. #### Responsibilities: - Establish and maintain socket connection with the Rhino Plugin - Format commands as JSON messages - Send commands to the Rhino Plugin - Receive and parse responses - Provide a clean API for the MCP Server ### 3. MCP Server (`src/rhino_mcp/mcp_server.py`) The MCP Server implements the Model Context Protocol and exposes Rhino functionality as MCP tools. #### Responsibilities: - Implement WebSocket server for MCP communication - Register available Rhino tools - Validate tool parameters - Forward tool invocations to the Rhino Client - Format responses according to MCP specifications ## Data Flow 1. **MCP Request Flow**: - Claude AI sends a tool invocation request to the MCP Server - MCP Server validates the request and parameters - MCP Server formats the request for the Rhino Client - Rhino Client sends the request to the Rhino Plugin - Rhino Plugin executes the requested operation in Rhino - Results flow back through the same path 2. **Communication Formats**: - MCP Server <-> Claude AI: JSON-RPC over WebSockets - Rhino Client <-> Rhino Plugin: Custom JSON protocol over TCP sockets ## Error Handling The system implements multiple levels of error handling: 1. **MCP Server**: Validates requests and parameters before forwarding 2. **Rhino Client**: Handles connection issues and timeouts 3. **Rhino Plugin**: Catches exceptions during command execution in Rhino 4. **All Components**: Provide detailed error messages with stack traces for debugging ## Extensibility The architecture is designed for extensibility: 1. **Tool Registration**: New tools can be added to the MCP Server without modifying the core code 2. **Command Handlers**: The Rhino Plugin can be extended with new command handlers 3. **Protocol Versioning**: Both socket protocols include version information for compatibility ## Security Considerations 1. **Local Connections Only**: Both socket servers bind to localhost by default 2. **No Authentication**: The current implementation assumes a trusted local environment 3. **Input Validation**: All component interfaces validate input to prevent injection attacks ## Future Considerations 1. **Geometry Transfer**: Optimize large geometry data transfer between components 2. **Connection Recovery**: Improve automatic reconnection for better resilience 3. **Tool Expansion**: Add more Rhino operations as MCP tools 4. **Authentication**: Add authentication for non-local deployments