<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> This project provides an AI-powered workbench for FreeCAD, enabling natural language 3D model generation. It targets FreeCAD users, engineers, and researchers, aiming to streamline complex modeling tasks through intelligent automation and direct LLM integration.

How It Works

The workbench integrates LLMs via a chat interface with "Plan" (code review) and "Act" (execution) modes. "Act" mode prioritizes 50+ structured FreeCAD operations for safer modeling, falling back to LLM-generated Python code. Features include tool reranking, reusable "skills" for complex tasks, user-defined extension tools, vision routing, and LLM-based error self-correction.

Quick Start & Requirements

Install by cloning the repository into FreeCAD's Mod directory and restarting. Requires FreeCAD 1.0+ (tested 1.0.2, 1.1.0) and an LLM provider (local Ollama or cloud API key). Configuration is managed via FreeCAD preferences or the workbench settings dialog.

Highlighted Details

Supports over 20 LLM providers (Anthropic, OpenAI, Ollama, Gemini, etc.) and custom OpenAI-compatible endpoints. Advanced features include context compacting, session resume, AGENTS.md support for project-specific instructions, and an opt-in "Dangerous mode" for advanced users to bypass safety checks. Adheres to a strict zero-external-dependencies policy, using hooks/MCP servers for features like PDF/DOCX conversion.

Maintenance & Community

Labeled "Alpha software," this project is in early development, warning of potential instability, breaking changes, and FreeCAD crashes. Contributions are welcomed. No specific community links were found.

Licensing & Compatibility

Code is LGPL-2.1 (copyleft); icons are CC0-1.0 (public domain). LGPL-2.1 may impose obligations on derivative works, potentially impacting closed-source commercial use.

Limitations & Caveats

As alpha software, expect rough edges and instability. Zero-external-dependency policy requires user-provided solutions for advanced file parsing. "Dangerous mode" carries significant risks of arbitrary code execution. Platform support is primarily Linux-tested; macOS/Windows behavior requires verification.