Forgent3D
Local AI CAD companion that pairs coding agents with a desktop viewer so they can generate, rebuild, screenshot, and verify parametric models and MJCF assemblies against real geometry instead of text-only reasoning.
What it does and how it connects
Forgent3D is a local AI CAD companion that pairs coding agents with a desktop viewer so they can generate, rebuild, screenshot and verify parametric models and MJCF assemblies against real geometry instead of text-only reasoning. The typical loop: ask the agent to create or modify a model, the agent edits part.py, asm.xml and params.json, calls the viewer rebuild tool, Forgent3D updates the preview and caches geometry info, and the agent uses screenshots or bounding-box data to verify the result. Models are build123d or OpenSCAD parts composed into MJCF assemblies with reusable STL meshes, joints, constraints and optional MuJoCo simulation.
The desktop app is in maintenance mode: bugs still get fixed, but no new features are coming. The README recommends forgent3d-skills instead, a lighter option with no 300-plus MB app to install: one command adds a skill to the coding agent the user already uses (Claude Code, Codex, Cursor and others), and the agent can then build, preview and measure build123d or OpenSCAD parts and export them to STEP, STL and 3MF. The release app is packaged with the local CAD runtime and viewer.
Alternatives
Compare side by side- Chamfer
A text and image driven CAD agent with a built-in verification gate, rather than a companion viewer that verifies an agent's models.
- MAC (Multi-Agent CAD)
Generates printable build123d parts through a multi-agent pipeline, rather than pairing a coding agent with a viewer.
- Text23D Mechanical
Turns chat prompts into CadQuery or FreeCAD models in a web workbench, rather than pairing a coding agent with a desktop viewer.
- vibecad (JoeMattie)
A Claude Code skill that edits build123d models through a browser canvas, rather than a desktop viewer companion.
- cad-agent (kerryshi)
Turns English into CadQuery through a frozen spec with spec-derived verification, rather than a viewer that verifies an agent's models.
Sources
Not verified
- Pricing: none stated on the repo.
- Real user voices: no real linked post read for this page; recent repo activity is the maintainers' own pull requests.
Last verified November 6, 2026. Listings are informational. The evidence score measures public, verifiable evidence and is not an endorsement or a compatibility claim.