Process point clouds (downsample, reconstruct surfaces, segment) with AI-suggested parameters tuned to your loaded model
Edit 3D meshes (simplify, smooth, fill holes, remesh) and get AI quality reviews on the output topology
Generate a 3D model from a text description and import it directly into your geometry processing workspace
Run advanced CGAL-backed algorithms (skeletonization, UV mapping, shape approximation) with AI flagging quality problems
Prebuilt binaries on GitHub Releases for a first look, building from source requires C++17, CMake, and optionally CGAL.
3D Claw is a desktop application for working with 3D geometry: loading models, running processing algorithms on them, generating new models from text descriptions, and inspecting results. It is built on top of an open-source library called Easy3D, which handles rendering, file reading, and the interactive viewer. What 3D Claw adds is a layer of AI participation throughout the workflow. The central idea is that AI should be available at every step of a geometry task, not just as a separate chat window. Before you run an algorithm, you can ask for parameter suggestions that take into account the specific model you have loaded and its measured properties. While a long-running algorithm is executing, the viewport can show live previews and intermediate states rather than showing nothing until the algorithm finishes. After the algorithm completes, the AI can review the output statistics and flag quality problems or suggest a next step. The same pattern applies to generating 3D models from a text prompt: the result imports directly into the workspace as a normal model you can then process further. The application supports a range of geometry operations grouped roughly into point-cloud processing, surface-mesh editing, and more advanced algorithms that require an optional library called CGAL. Point-cloud operations include things like downsampling, surface reconstruction, and segmentation. Mesh operations include simplification (reducing triangle count), smoothing, hole filling, and remeshing. The more advanced CGAL-backed tools cover skeletonization, deformation, UV mapping, and shape approximation. A health report panel shows topology statistics and flags common problems. Prebuilt binaries for Windows and Linux are available on the GitHub Releases page, which avoids the need to compile from source for a first look. The interface uses a panel-and-dock layout with a model tree, property inspector, history, measurements, and display controls. The application runs on Windows, Linux, and macOS and is licensed under the GPL. Building from source requires a C++17 compiler and CMake. CGAL is optional: without it, viewing, file loading, AI chat, and a subset of algorithms still work.
← tangtangtang1995 on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.