Text-to-CAD
Also known as: text to CAD, prompt-to-CAD, natural language CAD
CAD (computer-aided design) is the software mechanical engineers, product designers, and architects use to create precise 3D models of physical objects. Traditionally it requires specialized skill: you learn the tool, you manually define geometry, dimensions, and constraints. Text-to-CAD applies the same prompt-to-output pattern that text-to-image brought to visual art, but the output is a functional, editable 3D model rather than a picture.
The distinction that matters is parametric versus mesh. A mesh is a surface approximation, useful for rendering but not for manufacturing. A parametric model (the kind engineers actually work with) stores the design intent: this hole is 8mm in diameter, this wall is 3mm thick, this fillet has a 2mm radius. The best text-to-CAD tools in 2026 produce parametric models that open in SOLIDWORKS, Fusion 360, or CATIA and stay editable. Tools like Zoo Design Studio and MecAgent are leading this space, and the earthtojake/text-to-cad open-source repo became a popular sandbox for agent-driven CAD generation.
For the builder community, text-to-CAD matters as a signal that generative AI is expanding into physical-world domains. Anything that connects AI output to manufacturing, robotics, or hardware prototyping will need some version of this capability. Agent-driven CAD, where an AI agent iterates on a design based on constraints or test results, is the natural next step.