Fusion 360 MCP Server

What is Fusion 360 MCP Server?

The Fusion 360 MCP Server is a Model Context Protocol (MCP) server that interfaces between Cline and Autodesk Fusion 360, exposing Fusion's toolbar-level commands as callable tools that map directly to the Fusion API. It is intended for users who want to control Fusion 360 via natural language prompts processed by Cline, with the server generating Python scripts executable in Fusion 360.

How to use Fusion 360 MCP Server?

Install Python 3.9+, clone the repository, and install dependencies. Run python main.py for HTTP server mode (port 8000) or python main.py --mcp for MCP stdin/stdout mode. For Cline integration, add the server configuration to your MCP settings file with the --mcp flag. Available endpoints include GET /tools and POST /call_tool/POST /call_tools for single or sequential tool calls.

Key features of Fusion 360 MCP Server

• Exposes Fusion 360 toolbar commands as callable MCP tools
• Parses natural language prompts into Fusion tool actions
• Generates executable Python scripts for Fusion 360
• Supports HTTP API and MCP stdin/stdout modes
• Tool definitions stored in src/tool_registry.json
• Extensible with new tools via registry and script templates

Use cases of Fusion 360 MCP Server

• Automating 3D modeling tasks via natural language instructions
• Generating Fusion 360 scripts for repeated design operations
• Integrating Fusion 360 with AI assistants like Cline
• Prototyping parametric models without manual UI interaction

FAQ from Fusion 360 MCP Server

What does Fusion 360 MCP Server do differently from the Fusion 360 API?

It acts as a bridge between Cline and Fusion 360, mapping high-level tool commands (e.g., Extrude, Fillet) to the Fusion API through MCP, and returns ready-to-run Python scripts rather than requiring direct API calls.

What are the runtime requirements?

Python 3.9 or higher and Autodesk Fusion 360 must be installed. The server runs as a local process; no cloud dependencies are mentioned.

Where are tool definitions stored?

All tool definitions (name, description, parameters, docs) are stored in src/tool_registry.json. The server reads this file to determine available endpoints and parameter schemas.

How does the server transport data?

It supports two transports: an HTTP server (FastAPI on port 8000) and MCP standard I/O (stdin/stdout) mode, triggered by the --mcp flag.

Is authentication or authorization required?

No authentication or authorization mechanisms are described in the README. The server appears to run locally without built-in security controls.