CAD-MCP Server

Overview

CAD‑MCP is a Model Context Protocol (MCP) server that bridges natural language interfaces with desktop CAD applications. It empowers AI assistants—such as Claude—to issue real‑world drawing commands to AutoCAD, GstarCAD, and ZWCAD without the user needing to manually operate any UI. By interpreting plain text instructions into precise CAD operations, the server transforms conversational design workflows into tangible geometry.

The core problem it solves is the disconnect between AI’s text‑based reasoning and the command‑driven nature of legacy CAD software. Traditional CAD workflows require specialized knowledge of menus, shortcuts, and scripting languages. CAD‑MCP removes this barrier: a user can simply say “draw a 10 mm circle at (5,5) on layer ‘Detail’” and the server will translate that into the appropriate COM calls, create the circle, set its layer, and apply any color or annotation requested. This reduces onboarding time for engineers, architects, and designers who prefer conversational interfaces.

Key capabilities include:

• Multi‑software support: Works with AutoCAD, GstarCAD (GCAD), and ZWCAD, handling each system’s quirks through a unified API.
• Comprehensive drawing primitives: Lines, circles, arcs, rectangles, polylines, text, hatch patterns, and dimensions are all exposed as MCP methods.
• Natural language parsing: The server’s NLP engine interprets shape, action, and color keywords from user sentences, mapping them to CAD parameters.
• Layer management: Users can create and switch layers directly via text, enabling organized drawings.
• File output: The drawing can be saved automatically to a specified directory, making the process fully automated.

In real‑world scenarios, CAD‑MCP can be integrated into design review tools where an AI assistant provides instant visual feedback on textual specifications, or into educational platforms that teach CAD concepts through guided dialogue. It also supports rapid prototyping pipelines where a product manager’s feature list is converted into CAD models by an AI orchestrator.

Integration with existing MCP workflows is straightforward: developers expose the server in their or use the MCP Inspector for debugging. Once running, any AI client that understands MCP can call methods like or , receiving immediate visual results in the target CAD application. The server’s design emphasizes low latency, reliable COM communication, and a clean configuration file that lets teams adjust startup waits or command delays to match their hardware.

Overall, CAD‑MCP turns conversational intent into executable CAD actions, streamlining design creation and opening the door for AI‑driven drafting in environments that have traditionally been resistant to automation.