How to Install the text-to-cad Library: A Complete Setup Guide

Install the text-to-cad library using the Skills CLI by running npx skills add earthtojake/text-to-cad, which registers all CAD agent skills with your runtime.

The text-to-cad repository by earthtojake is not a traditional Python package but a collection of agent skills that expose CAD generation, robot-description, and fabrication functionality. Instead of pip install, you add these skills to your agent runtime—whether that's Skills CLI, Codex, Claude Code, or Grok Build—to enable natural-language CAD generation directly within your development environment.

Installation Methods

The text-to-cad library supports multiple installation paths depending on your agent runtime. All methods pull from the same source repository but register the skills through different plugin systems.

Option 1: Skills CLI (Recommended)

The Skills command-line tool is the preferred installation method. It fetches the repository, creates the appropriate skill directories under skills/, and registers them with any agent that supports the Skills system.

npx skills add earthtojake/text-to-cad

This command installs all current skills from the repository. To update existing skills or add newly introduced ones, simply re-run the same command. The alias npx skills install earthtojake/text-to-cad also works. According to the source code in README.md (lines 92-99), this is the primary installation mechanism documented for general use.

Option 2: Codex Plugin

For OpenAI Codex agent users (version 0.142.0 or later), install via the native plugin marketplace:

codex plugin marketplace add earthtojake/text-to-cad
codex plugin add cad@text-to-cad

Option 3: Claude Code Plugin

If you are using Claude Code, register the repository through its plugin system:

claude plugin marketplace add earthtojake/text-to-cad
claude plugin install cad@text-to-cad

Option 4: Grok Build Plugin

For Grok Build environments, install with trust verification:

grok plugin install earthtojake/text-to-cad --trust
grok plugin enable cad

These plugin-based installations are documented in README.md lines 117-141, which detail how each agent runtime registers the same underlying skill set.

What Gets Installed

Understanding the installation footprint helps troubleshoot issues and extend functionality.

Skill Directory Structure

Each skill resides under the skills/ directory in the repository. The primary components include:

• skills/cad/SKILL.md – Defines the core CAD generation skill and its capabilities
• skills/cad-viewer/SKILL.md – Provides the CAD Viewer for previewing generated geometry
• Individual skill implementations that depend on the cadgen distribution

Python Dependencies

The installation automatically pulls the cadgen package from PyPI, which serves as the core CAD kernel. Each skill's requirements.txt pins a specific compatible version of cadgen, ensuring that the Python runtime has access to the underlying geometry engine required for STEP file generation and mesh operations.

JavaScript Runtime

The CAD Viewer component relies on cadgen-js, documented in packages/cadgen-js/README.md. This JavaScript runtime installs automatically when you add the viewer skill, enabling web-based preview of generated models without manual configuration.

Verifying Your Installation

After installation, confirm that the skills are properly registered and functional.

First, list installed skills to verify registration:

npx skills list

You should see entries for cad, cad-viewer, and other text-to-cad skills.

Next, test the CAD generation capability:

cadgen cad --help
cadgen cad "a simple box 100mm wide"

The second command generates a STEP file and, if the viewer skill is active, launches the CAD Viewer for visual validation. Successful execution confirms that both the skill registration and the cadgen Python backend referenced in skills/cad/SKILL.md are operational.

Updating the Library

Since text-to-cad is distributed as living agent skills rather than a version-locked package, updates are applied by re-running your original installation command:

npx skills add earthtojake/text-to-cad

This fetches the latest skill definitions from the main branch and updates the cadgen dependency if the requirements.txt has changed. For plugin-based installations, use your agent's native update command (typically plugin update or plugin add with the same identifier).

Summary

• Install the text-to-cad library via npx skills add earthtojake/text-to-cad for universal compatibility with Skills-compliant agents
• Alternative installations use native plugin commands for Codex (≥0.142.0), Claude Code, and Grok Build
• The installation creates skill directories under skills/ and automatically installs the cadgen Python package and cadgen-js runtime
• Verify functionality by running npx skills list and generating a test STEP file with cadgen cad
• Update existing installations by re-running the installation command to pull the latest skill definitions

Frequently Asked Questions

Is text-to-cad available on PyPI?

No. The text-to-cad library is not distributed as a standalone Python package on PyPI. Instead, it is a collection of agent skills that automatically manage their Python dependencies (including the cadgen kernel) through the skill installation process defined in the repository's README.md.

Can I use text-to-cad without an AI agent like Codex or Claude Code?

Yes. The Skills CLI (npx skills) functions independently of specific AI agents. You can install the skills using the CLI and then invoke them via the cadgen command-line interface directly, making the CAD generation capabilities available in standard shell environments without requiring an AI coding assistant.

Which version of Codex supports the text-to-cad plugin?

Codex version 0.142.0 or later is required to support the plugin marketplace commands needed to install text-to-cad. Earlier versions of Codex lack the plugin marketplace subcommand interface necessary to register external skill repositories.

How do I uninstall the text-to-cad library?

To remove the skills, use the reverse of your installation method: run npx skills remove earthtojake/text-to-cad for CLI installations, or use codex plugin remove cad@text-to-cad, claude plugin uninstall cad@text-to-cad, or grok plugin disable cad depending on your agent runtime configuration.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}