cadquery-mcp 0.1.0

MCP (Model Context Protocol) server for CadQuery - enables AI assistants to create 3D CAD models

CadQuery MCP Server

An MCP (Model Context Protocol) server that enables AI assistants like Claude to execute CadQuery scripts and render 3D CAD models.

Features

• render - Execute CadQuery code and return SVG images of the 3D model

  • Multiple camera angles: isometric, front, back, top, bottom, left, right
  • Multi-view mode for complex models
  • Configurable image dimensions
  • Hidden line rendering

• inspect - Get geometry information about a shape

  • Bounding box dimensions
  • Volume and surface area
  • Center of mass
  • Topology counts (solids, faces, edges, vertices)

• get_parameters - Extract customizable parameters from CadQuery scripts

• export - Export models to various formats

  • STEP, STL, SVG, DXF, AMF, 3MF, VRML, BREP

Installation

Prerequisites

CadQuery must be installed first. The recommended method is via conda:

conda install -c conda-forge cadquery

Install from Source

git clone https://github.com/CadQuery/cadquery-contrib.git
cd cadquery-contrib/mcp-server
pip install .

For development (editable install):

pip install -e .

Run Tests

pip install pytest
pytest test_cadquery_mcp_server.py -v

Configuration

Claude Code

Add to your ~/.claude/settings.json:

{
    "mcpServers": {
        "cadquery": {
            "command": "cadquery-mcp"
        }
    }
}

Claude Desktop

Add to your Claude Desktop configuration (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
    "mcpServers": {
        "cadquery": {
            "command": "cadquery-mcp"
        }
    }
}

Note: If using conda, you may need to specify the full path:

{
    "mcpServers": {
        "cadquery": {
            "command": "/path/to/conda/envs/yourenv/bin/cadquery-mcp"
        }
    }
}

Usage Examples

Once configured, you can ask Claude to create 3D models:

"Create a box with a hole through it"

Claude will execute:

import cadquery as cq

result = (
    cq.Workplane('XY')
    .box(20, 20, 10)
    .faces('>Z')
    .workplane()
    .hole(5)
)

And return a rendered SVG image of the model.

Multi-View Rendering

For complex models, request multiple views:

"Show me this bracket from multiple angles"

The server will return isometric, front, top, and right views.

Parametric Models

CadQuery scripts can define parameters:

height = 10.0  # Height of the box
width = 20.0   # Width of the box
depth = 5.0    # Depth of the box

import cadquery as cq
result = cq.Workplane('XY').box(width, height, depth)

Use the get_parameters tool to extract these for modification.

Exporting Models

Export to STEP for manufacturing or STL for 3D printing:

"Export this model as a STEP file to ~/models/bracket.step"

API Reference

render

Execute CadQuery code and return rendered image(s).

Parameter	Type	Default	Description
code	string	required	CadQuery Python code to execute
view	string	"isometric"	Camera angle (isometric, front, back, top, bottom, left, right, isometric_back)
multi_view	boolean	false	Return multiple views
width	integer	800	Image width in pixels
height	integer	600	Image height in pixels
show_hidden	boolean	true	Show hidden lines

inspect

Get geometry information about the resulting shape.

Parameter	Type	Default	Description
code	string	required	CadQuery Python code to execute

get_parameters

Extract customizable parameters from a script.

Parameter	Type	Default	Description
code	string	required	CadQuery Python code to parse

export

Export the model to a file.

Parameter	Type	Default	Description
code	string	required	CadQuery Python code to execute
filename	string	required	Output filename
format	string	auto	Export format (STEP, STL, SVG, DXF, AMF, 3MF, VRML, BREP)

Writing CadQuery Scripts for MCP

Scripts should either:

1. Assign the final shape to a variable named result:

   result = cq.Workplane('XY').box(1, 2, 3)
   

2. Use show_object() to output shapes:

   box = cq.Workplane('XY').box(1, 2, 3)
   show_object(box)
   

License

Apache License 2.0 - see LICENSE for details.

Contributing

Contributions are welcome! Please see the cadquery-contrib repository for guidelines.