CAD is where generative AI usually falls apart — geometry needs precision, and a part that’s 95% right is a part that doesn’t fit. text-to-cad, an open-source project from earthtojake, is the first attempt in a while that hardware people are actually passing around. 9.8k stars, 600+ in a single day, trending on GitHub.
What it actually is
Not a chat app. It’s a harness — a collection of agent skills that plug into Claude Code and Codex. You describe a part, and the agent writes source-controlled CAD scripts (build123d on top of OpenCascade) that compile into parametric, editable models. Output goes to STEP, STL, 3MF, and GLB. It also pulls off-the-shelf STEP parts like screws, bearings, and motors, generates 2D DXF drawings, and writes URDF robot files with MoveIt planning groups.
Why it matters
It’s open source — self-host it, or install the plugin into a coding agent you already run. The bet is smart: instead of generating geometry directly, it makes the agent write code that generates geometry. Text-to-CAD as text-to-code. That’s why the output is editable instead of a black-box mesh — and why the robotics crowd thinks this one might stick.
You Might Also Like
- Heretic Just hit Github Trending and the ai World has Opinions
- Pentagi Just hit 1 on Github Trending and Yeah its Worth the Hype
- Pageindex Just hit Github Trending and it Might Make you Rethink rag Entirely
- Openfang Just Dropped and its Already the Hottest Agent os on Github
- Shuo sub 500ms Voice Agent 600 Lines of Python That Make Voice ai Feel Instant

Leave a comment