vcad

vcad is a powerful, open-source parametric CAD system designed specifically for the AI era, providing a robust suite of tools for agents to programmatically generate and manipulate 3D geometry. By integrating directly with AI workflows through the Model Context Protocol (MCP), vcad enables developers to automate the creation of mechanical components, brackets, plates with hole patterns, and complex assemblies. The system operates on a professional-grade Rust-based BRep (Boundary Representation) kernel, ensuring high-fidelity mathematical accuracy for engineering and manufacturing tasks. Users can define geometry using a structured primitive-based approach—including cubes, cylinders, spheres, and cones—and perform advanced boolean operations like unions, differences, and intersections, or apply parametric patterns such as linear or circular repetitions.

• Full support for parametric design, allowing for dynamic scaling, hole placement, and complex geometry definition via JSON-based tool calls.

• Advanced modeling features including fillets, chamfers, shell operations, sketching with 2D constraints, extrude, revolve, sweep, and loft functions.

• Seamless integration with AI agents through the vcad MCP server, providing standard tools like create_cad_document, export_cad, and inspect_cad for geometric validation.

• Versatile import and export capabilities, supporting industry-standard formats including STL, GLB, STEP, and DXF for 3D printing, visualization, and CAD interoperability.

• High-performance simulation support with physics integration via Rapier3D, allowing agents to design mechs and test articulation or kinematics within a gym-style interface.

• Built-in rendering capabilities for creating isometric drafting-style diagrams, useful for technical documentation and design visualization.

• The system is ideal for engineers, robotics developers, and AI researchers who need to bridge the gap between generative AI and physical manufacturing.

• Practical workflows include building custom mounting plates, enclosures, mechanical bushings, and complex rail systems for robotics applications.

• Input for the system consists of structured JSON commands that specify geometric parameters, positioning (absolute, named, or percentage-based), and sequence of operations.

• Output artifacts are typically validated via inspect_cad to confirm volume and bounding boxes before being exported to STL or GLB for physical production or rendering.

• Users should note that primitive origins vary by shape—cubes at corners and cylinders at base centers—and coordinate systems are optimized for precision at the millimeter scale.