Adam Launches CADAM: Open-Source Text-to-CAD Platform Runs Entirely in Browser
Key Takeaways
- ▸Adam launches CADAM, an open-source Text-to-CAD platform that generates parametric 3D models from natural language descriptions and images
- ▸Browser-based execution via WebAssembly eliminates installation requirements while supporting interactive parameter adjustment without re-running the AI model
- ▸Model-agnostic architecture supports Claude, Gemini, and OpenAI, with Gemini 3.1 Pro currently showing top performance in the team's evaluations
Summary
Adam, a Y Combinator W25 startup, has launched CADAM, an open-source AI-powered platform that transforms natural language descriptions and images into parametric 3D CAD models. Built on React and Supabase, CADAM generates OpenSCAD code that can be exported as .STL, .SCAD, .DXF, and other formats, positioning itself as an "AI TinkerCAD" for mechanical design.
The platform operates through an agentic architecture that uses either parametric mode (generating OpenSCAD code) or mesh mode (creating 3D textured meshes) by swapping system prompts and tools. Users can adjust generated models through interactive sliders that trigger deterministic parameter updates without requiring additional LLM calls. CADAM runs entirely in-browser using WebAssembly compilation of OpenSCAD and 3D rendering via Three.js, eliminating installation requirements.
The platform is model-agnostic via the Vercel AI SDK, supporting Claude (Anthropic), Gemini (Google), and OpenAI models through OpenRouter, with adaptive thinking auto-enabled on newer models. According to the team's evaluations, Google's Gemini 3.1 Pro currently outperforms other models for CAD generation. CADAM supports BOSL, BOSL2, and MCAD libraries, plus custom font support, and is available as open-source code that users can clone and run locally.
- Open-source release democratizes AI-powered CAD and invites community contributions for future improvements like build123d and CadQuery support
Editorial Opinion
Adam's CADAM represents a significant leap forward in democratizing mechanical design through AI. By combining open-source accessibility with an intelligent agent architecture, the platform could fundamentally change how engineers and makers approach CAD—much as AI coding assistants have transformed software development. The browser-based implementation is particularly clever, suggesting that complex 3D design workflows no longer require expensive local software installations. If the quality and consistency of generated designs continue to improve, this could accelerate adoption of AI-assisted engineering across industries from hardware startups to established manufacturing firms.
