Open-source multi-agent framework that generates editable CAD parts and assemblies from text prompts using build123d.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
MAC (Multi-Agent CAD) is a decoupled multi-agent framework for text-to-CAD generation, developed by the IEI Lab at Tsinghua University (College AI). It turns a natural-language design brief into editable engineering geometry built on the build123d Python CAD library, rather than image-only output. It can generate a single printable part, or decompose a mechanism into separately generated components, assemble them, validate the result and export artifacts for inspection and simulation.
The v2 release adds visual verification, an independent Judge agent, and an experimental multi-part assembly workflow (marked a technology preview). Assemblies can be exported as URDF files for robotics and physics simulation; the README demonstrates a generated hand performing object rotation and pick-and-place in simulation. The authors state that the assembly examples are curated and are not a measured success rate, and that generated geometry and joints should be validated before manufacturing or simulation use.
On a ten-prompt, 141-feature single-part benchmark, the README reports MAC v1 with Qwen 3.7 at 140/141 features passed using 0.90M tokens (about $1.43), versus a reproduced CAD Skill baseline at 138/141 using 103.95M tokens (about $18.59). That is the source of the 116x fewer tokens and 13x lower cost claim. MAC v2 with Qwen 3.8 reports 140/141 at 1.40M tokens. These are the authors' own figures. The code is MIT licensed Python and is run by the user with their own model backend.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Multi-agent system that turns text prompts into printable CAD files. Claims 116x fewer tokens and 13x lower cost than CAD Skills with a higher pass rate; model agnostic and runs locally.
Closest neighbours by meaning, not just by category.
Research · Tencent
Tencent Hunyuan research framework that turns one text prompt into an explorable, editable open-world 3D scene.
Library
Blender, the free open-source 3D suite, driven by AI agents such as Claude through an MCP server.
Model · Stability AI
Research method adding typed 3D constraint meshes (hull, avoidance, touch) as control for text-to-3D generation on TRELLIS.2.
Model · Tencent
Tencent's unified multimodal model for 3D understanding, text-to-3D generation, instruction-based editing and part generation.
Model · InternLM
Multi-codebook diffusion LLM from Shanghai AI Laboratory for visual understanding, image generation and image editing.
Model · Tencent
Tencent's Hy4 multimodal model with visual understanding, computer use, 3D/Blender and knowledge-work abilities.
