text-to-cad

Product Introduction

1. Definition: Text-to-CAD is an open-source library of AI agent skills specifically engineered for the mechanical design and manufacturing workflow. It falls under the technical categories of Computer-Aided Design (CAD), Computer-Aided Engineering (CAE), and Computer-Aided Manufacturing (CAM) automation.
2. Core Value Proposition: It exists to solve the problem of automating complex, multi-step engineering workflows by providing AI agents with ready-to-use, focused skills. Its primary value is enabling developers and engineers to build AI agents that can interpret natural language or images and execute precise tasks like 3D model creation, robot file generation, simulation setup, and fabrication preparation, thereby bridging the gap between high-level intent and low-level engineering file output.

Main Features

1. CAD Skill (skills/cad): This core skill enables AI agents to create and edit parametric 3D CAD models from plain-language descriptions or image inputs. It utilizes a programmatic CAD kernel to generate geometry, with the Standard for the Exchange of Product Data (STEP) file format as the primary, industry-standard output. The skill also includes functionality for exporting to common formats like STL for 3D printing, 3MF for additive manufacturing, and GLB for web visualization.
2. URDF & SDF Skills (skills/urdf, skills/sdf): These skills allow AI agents to author robot description files critical for simulation and control. The URDF (Unified Robot Description Format) skill automates the creation of robot kinematic structures, defining links, joints, limits, inertial properties, and visual meshes. The SDF (Simulation Description Format) skill builds upon this for physics-based simulators like Gazebo, adding models, worlds, sensors, lights, and detailed physical properties.
3. Fabrication Pipeline Skills (skills/dxf, skills/sendcutsend, skills/gcode): This suite of skills automates the path from design to physical part. The DXF skill generates 2D vector drawings for laser cutting, CNC machining, or gasket creation. The SendCutSend skill performs pre-flight checks on DXF and STEP files before upload to the fabrication service. The G-code skill integrates real slicer CLIs (like PrusaSlicer or OrcaSlicer) to slice 3D meshes into validated, printer-profile-specific toolpaths for Fused Deposition Modeling (FDM) 3D printing.

Problems Solved

1. Pain Point: The engineering design process is highly fragmented, requiring expertise in multiple specialized software tools (CAD, CAE, CAM) and file formats. Manually translating a concept into a manufacturable design, a simulated robot, or a print-ready file is time-consuming, error-prone, and requires significant repetitive work.
2. Target Audience: The primary users are Developers building AI agents for engineering and Mechanical Engineers/Robotics Engineers seeking workflow automation. Secondary users include makers, hobbyists, and educators in robotics and digital fabrication who want to accelerate prototyping.
3. Use Cases: Essential scenarios include: an engineer using an AI agent to instantly generate a 3D bracket model from a verbal description; a robotics team automating the creation and validation of URDF files for a new robot arm; a startup quickly iterating on enclosure designs and sending validated DXF files directly to a laser cutting service; a user preparing a 3D model for printing by having an agent automatically slice it with optimal, printer-specific settings.

Unique Advantages

1. Differentiation: Unlike monolithic CAD software with built-in automation or cloud-based generative design services, Text-to-CAD is a modular, locally-runnable library of skills. It doesn't lock users into a specific CAD platform or cloud ecosystem. It directly compares to manual workflow scripting by providing pre-built, agent-native skills that are more accessible and composable than writing low-level API code for each individual tool.
2. Key Innovation: Its core innovation is the "skill" abstraction for engineering tasks. By packaging deep domain expertise (e.g., URDF syntax, DXF standards, slicer parameters) into discrete, AI-agent-callable functions, it allows general-purpose AI models to perform specialized engineering work without retraining. The commitment to being 100% open-source and running locally also addresses critical concerns for IP-sensitive industries regarding data privacy and vendor lock-in.

Frequently Asked Questions (FAQ)

1. What is Text-to-CAD and how does it work? Text-to-CAD is an open-source library that gives AI agents the ability to perform engineering tasks. It works by providing pre-programmed "skills"—focused modules for CAD modeling, robot file generation, and fabrication prep—that an AI agent can call upon when a user makes a natural language request, translating intent into specific file outputs like STEP, URDF, or G-code.
2. Can Text-to-CAD generate files for 3D printing and CNC machining? Yes, Text-to-CAD directly supports both additive and subtractive manufacturing workflows. Its cad skill can export 3D models to STL and 3MF formats for 3D printing, while its dxf skill creates 2D DXF drawings for CNC machining or laser cutting, and its sendcutsend skill validates files for fabrication services.
3. Is Text-to-CAD a cloud service or local software? Text-to-CAD is designed primarily as a local library. The skills run on your local machine, ensuring that sensitive design data never leaves your environment. It is installed via an npm-based CLI (npx skills add...) or as a plugin for compatible AI agent platforms.
4. How does Text-to-CAD help with robotics development? It significantly accelerates robotics development by automating the creation of essential description files. AI agents can use the urdf skill to generate robot kinematic structures, the srdf skill to add MoveIt-specific semantic data for motion planning, and the sdf skill to build full simulation environments with physics and sensors.
5. What CAD file formats does Text-to-CAD support? The primary output for 3D CAD is the industry-standard STEP (ISO 10303) format. It also supports export to common formats including STL (stereolithography), 3MF (3D Manufacturing Format), and GLB (binary glTF) for visualization. For 2D geometry, it supports DXF (Drawing Exchange Format).