Text-to-CAD Open Source: AI-Generated 3D Robot Models Enter Practical Stage

From Text to Assemblable 3D Parts

3D CAD design has long been the domain of engineers—tools like Fusion 360 and SolidWorks have high learning barriers, parameter adjustments are tedious, and gear meshing and tolerance fitting require repeated trial and error. The Text-to-CAD open-source toolchain is changing this: describe requirements in natural language, and AI automatically generates editable parametric 3D models.

The tool surpassed 1,000 stars on GitHub in 3 days, uses the MIT license, supports direct invocation in Claude Code or Codex, and outputs in STEP, STL, GLB, DXF, and URDF formats.

Tool Stack and Workflow

The core toolchain includes:

• Text-to-CAD Engine: Based on Python + build123d, converts natural language instructions into parametric CAD code
• GPT 5.5 / Codex: Acts as the reasoning layer, understanding engineering intent and generating build123d-executable Python scripts
• Export Pipeline: Supports STEP (precision manufacturing), STL (3D printing), URDF (robotics simulation), GLB (visualization preview)

A typical complete workflow:

1. Describe requirements in Claude Code, e.g., “Design a 7-DOF robot arm with a parallel gripper at the end”
2. The Text-to-CAD engine generates build123d Python scripts defining joint dimensions, connections, and motion ranges
3. Scripts run and output STEP assembly files and URDF kinematics models
4. URDF can be directly used in ROS for kinematic verification
5. STEP files go to 3D printing or CNC machining

In actual testing, a developer completed a full 7-axis robot arm URDF model in Codex using this workflow, including functional kinematics and a custom GUI—what would have taken weeks stitching together 6+ tools with traditional toolchains.

Comparison with Alternatives

Dimension	Fusion 360 / SolidWorks	Text-to-CAD
Input Method	Manual modeling	Natural language description
Learning Curve	Months	Minutes to get started
Model Precision	Industrial grade	Parametric grade, needs verification
Editability	Native editing	Modify prompt and regenerate
Robotics Simulation	Requires additional plugins	Native URDF output
Cost	$500+/year subscription	Free and open source

Quick Start

# 1. Install dependencies
pip install build123d

# 2. Enable Text-to-CAD in Claude Code or Codex
# Open project directory and describe your design requirements

# 3. Example prompt
# "Design a set of meshing gears, module 2, tooth counts 20 and 40,
#  center distance 60mm, output STEP file"

# 4. Run the generated Python script
python generated_cad.py
# Output: gears.step, gears.stl

Cost and Application Boundaries

Suitable for:

• Rapid iteration of robot prototypes
• Parametric design for 3D printed parts
• Educational and small-scale manufacturing scenarios
• Design tasks requiring frequent dimension adjustments

Limitations:

• Limited capability for complex surfaces (e.g., car bodies, consumer electronics enclosures)
• Generated models require engineering verification (stress analysis, tolerance checks)
• Not suitable for industrial production drawings requiring strict GD&T annotations
• Ecosystem still early-stage; community tutorials and best practices are limited

Text-to-CAD won’t replace professional CAD engineers’ tools, but it significantly lowers the technical barrier to 3D design, enabling developers to validate mechanical concepts in minutes.

Primary Sources

• colin-hoang/text-to-cad (GitHub)
• X Community Test: 7-Axis Robot Arm URDF Generation
• build123d Documentation