Source: github.com/anteloc/ldraw-mecha/tree/v0.5.0
Given an LDraw LEGO model, the agent infers its mechanics from the .mpd source, studies each mechanism, and plans how its parts interact. It then writes Python for Blender that creates the rigs and runs the planned animations. The author's approach splits models into submodels and mechanisms, so the agent handles each one in isolation.
The repository includes a setup script for macOS, with Linux and Windows under WSL2 marked as work in progress. Setup installs Blender 5.2.3, LeoCAD 26.09, uv and bun locally, and requires git-lfs and Python 3.12 or later. Users must also add several environment variables to their shell profile, and a model selector serves pre-annotated models for inspection.
The author lists known limits. Complex mechanisms and large models take too long to animate, and agents can get lost in long sessions. Flexible parts such as hoses and cables animate poorly, smaller models produce visible defects, and the tooling has no physics such as gravity, weight or inertia.
Why it matters if you build agents
Developers building agents that work with CAD or 3D scene files can use the repository's annotation and rig-generation approach as a reference for turning source files into executable Blender scripts. The repo's own AGENTS.md and PROCESS.md documents show how it is meant to be driven by an agent.
Anyone adopting it should expect to run the Blender and LeoCAD setup locally and to review generated animations, since the author reports defects on weaker models and no physics simulation.
Get stories like this every weekday
One short email with the day's agent tooling news.