amagine-ai/Amagine3D

Amagine3D: From hardware requirements to editable 3D designs

What it solves

Amagine3D simplifies the process of creating editable 3D hardware designs from high-level requirements. It eliminates the manual effort of translating natural-language descriptions, reference images, and physical dimensions into complex CAD models, specifically focusing on printable enclosures and assembly structures for intelligent hardware.

How it works

The system uses a 3D-native Agent that manages a "semantic scene" representing the design state. The Agent operates in an autonomous inner loop: it reads the current model, plans changes, executes them using Python-based CAD tools (build123d and Open CASCADE), and runs automated checks for wall thickness, interference, and printability. If checks fail, the Agent uses the measurements to refine the design. Once a candidate is verified, it is committed as a new version. The final output is editable Python source code and industry-standard files like STEP, STL, and 3MF.

Who it’s for

Hardware engineers, product designers, and electronics enthusiasts who need to quickly prototype printable enclosures and mechanical structures around internal components.

Highlights

  • Editable Source: Generates parametric Python code (build123d) rather than static meshes, allowing for post-generation adjustments.
  • Automated Verification: Performs real geometry checks for collisions, operating clearances, and wall thickness.
  • Visual Perception: Can use image-capable LLMs to review generated 3D previews against reference images.
  • Multi-format Export: Supports genuine STEP files for professional CAD, as well as STL and 3MF for 3D printing.
  • Web-based Workbench: Includes a UI for chatting with the Agent, previewing models in 3D, and modifying parameters.

Related

  • Project
    Pan-Chera/Multi-Agent-CAD

    MAC (Multi‑Agent CAD) is an open‑source framework that turns natural‑language design briefs into editable CAD parts or full assemblies. It uses a LangGraph‑orchestrated multi‑agent pipeline (LLM planning, code generation, visual verification, and automatic repair via Aider) built on the build123d kernel. Outputs include STEP/STL/GLB for manufacturing and URDF for robotics simulation. The project provides a browser UI, detailed QA reports, and benchmark results showing >99% feature‑pass rates with dramatically lower token usage and cost compared to prior CAD‑Skill baselines.

  • Project
    HorizonRobotics/EmbodiedGen

    A generative 3D world engine that converts text and images into simulation-ready 3D assets and environments for embodied AI, deployable across multiple robotics simulators.

  • Project
    Adam-CAD/CADAM

    An open-source text-to-CAD web app that uses AI to transform natural language and images into parametric 3D models via OpenSCAD.

  • Project
    img2threejs/img2threejs

    ...

  • Project
    Shpigford/nurb

    nurb is a locally‑run Python/macOS app that lets you describe a 3‑D part in plain language, has your existing LLM write real B‑Rep CAD code, checks the model against 13 print‑ability rules, shows live updates in a three.js viewer, and exports print‑ready 3MF (or STL/STEP/GLB) with slicer settings embedded.