Fusion

You can control Autodesk Fusion 360 programmatically through an MCP server that bridges AI assistants to Fusion 360 actions. This setup lets you create parametric cubes and other modeling tasks from natural language commands, enabling automated CAD workflows.

How to use

Connect an MCP-compatible AI assistant to the Fusion MCP server to start issuing modeling commands. The server exposes two access methods: a local HTTP endpoint for MCP requests and a local stdio interface to run the MCP server directly.

• Start the Fusion Server to receive MCP requests. It listens on http://localhost:8000. This server translates MCP calls into actions that Fusion 360 can perform.

• Start the MCP server to expose the generate_cube tool to your AI assistant. The MCP server runs with a stdio transport by default and communicates with the Fusion Server to execute tasks inside Fusion 360.

• Use the generate_cube tool from your AI assistant to create parametric cubes in Fusion 360. The tool accepts edge lengths and triggers the LiveCube script inside Fusion 360 to perform the modeling operation.

How to install

Prerequisites you need to prepare before running the servers.

• Autodesk Fusion 360 (2023 or newer)
• Python 3.9+ with pip
• Autodesk Platform Services account with API access
• MCP-compatible AI assistant (such as Claude in Windsurf environments)

Install Python dependencies used by the MCP server and related components.

pip install -r requirements.txt

Set up environment variables with your APS credentials and Fusion activity information. Create a keys.env file with the required values.

APS_CLIENT_ID=your_client_id
APS_CLIENT_SECRET=your_client_secret
FUSION_ACTIVITY_ID=your_activity_id

Install the LiveCube script into Fusion 360 so you can run it from within the application.

- Open Fusion 360
- Scripts and Add-Ins (Shift+S)
- Click the green + and choose Add script
- Point to the LiveCube folder in the project
- Run LiveCube to start its internal HTTP server on port 18080

Start the Fusion Server to listen for MCP requests.

python fusion_server.py

This starts the server at http://localhost:8000

## 

Run the MCP server to expose the MCP tools to your AI assistant.

python fusion_mcp.py

## 

Configure your MCP client to connect to the Fusion MCP server by using an entry like this in your client configuration.

{ "mcpServers": { "fusion": { "command": "python", "args": ["fusion_mcp.py"] } } }

## Available tools

### generate\_cube

Creates a parametric cube in Fusion 360 by sending edge length parameters through the MCP interface to the LiveCube script, which runs inside Fusion 360 and exposes an HTTP endpoint for control.