The Idea.
CAD tools are designed for people who already know CAD. The bottleneck isn't the modeling — it's the translation from design intent to a software interaction model. Forge routes around that bottleneck entirely: describe what you want in plain English and let an LLM extract the geometric parameters.
AI Backend
Google Gemini API
3D Renderer
Three.js
Styling
Tailwind CSS
Export
STL Pipeline
Natural Language → Geometry.
"Create a 20mm cube with a 5mm hole through the center"
→
BoxGeometry(20,20,20) + CylinderGeometry(2.5, 2.5, 20) boolean subtract
Gemini extracts: size=20, hole_radius=2.5, orientation=Z
"A 50mm cylinder, 15mm tall, chamfered top edge"
→
CylinderGeometry(25, 25, 15) + edge chamfer modifier
Gemini extracts: r=25, h=15, edge_treatment=chamfer
"Mount bracket, two 4mm holes, 30mm apart"
→
Plate geometry + 2× CylinderGeometry(2,2) at x=±15
Gemini extracts: hole_count=2, hole_r=2, spacing=30
Key Features.
01
Intent Recognition via LLM
Gemini parses free-form geometry descriptions — extracting dimensions, shapes, relationships, and modifiers — and returns structured parameters the Three.js renderer can consume directly.
02
Live 3D Preview
Geometry renders in the browser in real time via Three.js. Orbit controls let users inspect from any angle before export. No install, no file management.
03
STL Export
Custom exporter converts Three.js BufferGeometry to manufacturing-ready STL files for FDM 3D printing. Output goes straight to a slicer.
Why Forge.
The gap Forge targets isn't between beginner and expert CAD users — it's between the moment of a design idea and the moment you can actually model it. Forge makes that gap zero. It's not a replacement for parametric CAD on complex assemblies; it's an accelerator for the first 80% of simple parts that currently require a 20-step tutorial just to start.