# text-to-cad

> Agent skills for CAD, CAE, CAM and robot description files

- Category: Multimodal Generation / 3D & CAD Generation
- Language: Python
- Licence: MIT
- Repository: https://github.com/earthtojake/text-to-cad
- Website: https://www.texttocad.dev
- Canonical page: https://ai-tldr.dev/tools/text-to-cad/

## In short

text-to-cad is a library of agent skills for generating, inspecting, sourcing, slicing and handing off CAD and robot-description artefacts from local project files. It sits in the 3D & CAD Generation category of the Multimodal Generation tools and is open-source, written in Python and licensed MIT. To get started: install with the skills CLI.

## Overview

text-to-cad is a library of agent skills for generating, inspecting, sourcing, slicing and handing off CAD and robot-description artefacts from local project files. Rather than one prompt that tries to do everything mechanical, it splits the domain into ten focused skills, each with its own instructions and its own Python tooling, so an agent asked for a bracket reaches for the CAD skill and an agent asked about print orientation reaches for the DfAM checker.

The CAD skill creates and edits models from plain-language or image requests, exporting STEP as the primary format with STL, 3MF and GLB alongside it. Around it sit skills for the rest of the workflow: CAD Viewer for local browser previews, step.parts for sourcing off-the-shelf screws, bearings, motors and connectors, DXF for 2D profiles and cut layouts, and G-code for slicing a mesh into a printer-profiled `.gcode` file through real slicer CLIs rather than an approximation.

Robotics gets first-class treatment, which is unusual for a text-to-CAD project. URDF writes robot structure — links, joints, limits, inertials, meshes — with SRDF adding MoveIt planning groups, end effectors, poses and collision rules, and SDF producing simulator models and worlds with frames, physics, sensors and lights. Two skills handle handoff to the physical world: SendCutSend validates DXF and STEP files before upload, and DfAM Check measures wall thickness, overhangs, support volume and build orientation per process. Python 3.11+, MIT licensed, with native plugin installs for Codex, Claude Code and Grok Build in addition to the skills CLI.

## What it does

- CAD skill: plain-language or image → editable models, exporting STEP plus STL, 3MF and GLB
- CAD Viewer for local browser previews of CAD and robot files without a separate viewer app
- step.parts sources real off-the-shelf components — screws, bearings, motors, connectors
- URDF, SRDF and SDF skills for robot structure, MoveIt planning groups and simulator worlds
- G-code slicing through real slicer CLIs, and DfAM Check for wall thickness, overhangs and build orientation
- SendCutSend validation of DXF and STEP files before you send them for fabrication
- Native plugins for Codex, Claude Code and Grok Build alongside the generic skills CLI

## Getting started

Python 3.11 or newer. The skills CLI path is the one the README recommends; provider-native plugins are available if you prefer them.

1. Install with the skills CLI

   This is the preferred path — it installs the individual skills directly for supported agents.

```bash
npx skills add earthtojake/text-to-cad
```

2. Or install as a provider-native plugin

```bash
# Codex (requires Codex 0.142.0 or newer)
codex plugin marketplace add earthtojake/text-to-cad
codex plugin add cad@text-to-cad

# Claude Code
claude plugin install cad@text-to-cad

# Grok Build
grok plugin install earthtojake/text-to-cad --trust
```

3. Update by re-running add, not update

   The README is explicit about this: `npx skills add` re-fetches the package and picks up skills added in a newer release, while `npx skills update` only refreshes what is already in your lockfile and silently misses new ones.

```bash
npx skills add earthtojake/text-to-cad
```

4. Remove a retired skill by hand

   Neither command removes a skill that was dropped upstream.

```bash
npx skills remove <skill>
```

## When to use it

- Describe a part in plain language and get a STEP file you can open in real CAD software
- Generate and sanity-check URDF/SRDF for a robot without hand-writing the XML
- Check a mesh for printability — wall thickness, overhangs, support volume — before committing a print
- Validate DXF and STEP files against a fabrication service's requirements before uploading them
