Fusion360

You can control Fusion 360 modeling tasks with natural language by running an MCP server that bridges your language model to Fusion 360 through a plugin. This setup lets you describe a part or feature in plain language and have it translated into modeling actions inside Fusion 360, delivering fast, AI-assisted design workflows.

How to use

You connect your language model client to the MCP server and issue modeling requests in natural language. The MCP server translates your requests into actions executed by the Fusion 360 plugin, which in turn calls the Fusion 360 Python API to perform the modeling work. You can ask for new documents, create shapes, edit features, insert parts from a library, or run arbitrary Python code inside Fusion 360. All communication ultimately happens on your local machine, with the Fusion 360 plugin listening on a fixed port and the MCP server handling JSON-RPC style calls from your MCP client.

How to install

Prerequisites you need before installation are Python 3.11 or newer, Fusion 360 (latest version), the FastMCP system, and access to a compatible LLM API such as OpenAI or Qwen.

Install the MCP server package using the recommended runtime tool for your environment. The project suggests using the universal executor to run the server directly from its distribution.

uvx --from fusion360-mcp fusion360_mcp --help

Configuration and runtime notes

The MCP server is invoked via a stdio-based protocol from your LLM client. The Fusion 360 plugin exposes a local HTTP API on port 9000 to receive requests from the MCP server. Ensure port 9000 is free and not used by any other application.

Set environment variables for your LLM and MCP server client to enable proper routing and authentication. The configuration typically includes your LLM API key, the LLM endpoint, and the MCP server URL where your LLM client will connect.

Troubleshooting and tips

If you encounter port conflicts, verify that port 9000 (Fusion 360 plugin HTTP service) is available and not bound by another process. Use port inspection commands to confirm usage and adjust as needed.

Check the health endpoints to confirm that both the MCP server and the Fusion 360 plugin are running correctly. The MCP client will connect to the MCP server via stdio, while the Fusion 360 plugin exposes its health endpoint on port 9000.

Example commands to run and test

Create a simple cylinder with a given height and diameter, or draw a square on the XY plane and extrude it. You can also create a gear with specific modulus and tooth count, or insert a part from the library. These examples illustrate the kinds of modeling tasks you can describe in natural language to trigger corresponding Fusion 360 actions.

API and endpoints overview

The MCP server exposes endpoints to create models, add features, modify parameters, and query model status. Use the MCP client to send commands like create, edit, and query to drive modeling sessions within Fusion 360.

Install steps (condensed)

1. Install prerequisites: Python 3.11+, Fusion 360, and FastMCP.

2. Install the MCP server package using the recommended runtime tool.

uvx --from fusion360-mcp fusion360_mcp --help

Available tools

create_document

Create a new Fusion 360 document to host models and assemblies.

create_object

Create a new object in the active Fusion 360 document.

edit_object

Modify properties or geometry of an existing object.

delete_object

Remove an object from the document.

execute_code

Run arbitrary Python code inside Fusion 360.

insert_part_from_library

Insert a part from the Fusion 360 parts library into the current document.

get_view

Capture and retrieve a screenshot of the current view.

get_objects

List all objects in the active document.

get_object

Retrieve a specific object by identifier.

get_parts_list

List parts available in the parts library.