9 MCP Servers for Computer-Aided Drafting (CAD) with AI

The Model Context Protocol (MCP) transforms how engineers and designers work with Computer Aided Drafting (CAD) software. Similar to how USB standardized hardware connections, MCP creates standardized connections between AI models and CAD applications (among other applications). This protocol enables LLMs like Claude, ChatGPT, and Gemini to control your drafting tools directly, automating everything from basic sketching to complex assembly modeling.

CAD professionals spend countless hours on repetitive tasks - creating standard components, generating drawing sheets, and updating design variations. MCP eliminates much of this drudgery by letting you describe what you want in plain English. Tell your AI assistant to, "create a mounting bracket with four bolt holes," and watch as it generates the parametric model automatically.

I've identified nine MCP servers specifically built for CAD integration, covering major platforms from AutoCAD to Fusion 360 to open-source FreeCAD. You can even plug into Onshape’s browser-based offering. These tools demonstrate how AI can augment engineering workflows (without replacing the engineer's creative problem-solving). If none of these nine meet your needs, you can also browse 6 MCP Servers for Using AI to Generate 3D Models, which accounts for tools like Revit, SketchUp, Blender, and even Thingiverse. 

AutoCAD control & automation

1. CAD-MCP by daobataotie (98 ⭐ on Github)

This comprehensive MCP server controls multiple CAD platforms through natural language commands. The project acts as a universal translator, converting conversational instructions into precise CAD operations across AutoCAD, GstarCAD, and ZWCAD environments.

Tools available

External APIs and technologies

• CAD Software COM Automation: Direct integration with AutoCAD, GstarCAD, and ZWCAD through Windows COM interfaces

• Python Libraries: pywin32 for Windows API access, pydantic for data validation, mcp for protocol implementation

• Natural Language Processing: Custom NLP module using regular expressions and keyword mapping for command interpretation

• Windows COM (Component Object Model): Core technology enabling Python applications to control Windows-based CAD software

Configuration requirements

• Primary configuration (src/config.json):

  • cad.type: Specifies target CAD application ("AUTOCAD", "GCAD", or "ZWCAD")

  • cad.startup_wait_time: Launch delay in seconds (default: 20)

  • cad.command_delay: Inter-command delay for reliability (default: 0.5)

  • output.directory: Default save location for drawings

  • output.default_filename: Standard naming convention for exports

The server requires Windows OS due to COM dependencies and needs one of the supported CAD applications installed locally. Integration with Claude Desktop involves modifying the claude_desktop_config.json file to specify the Python interpreter path and src/server.py location.

2. Easy-MCP-AutoCad by zh19980811 (64 ⭐ on Github)

Next up, another server specialized for AutoCAD, this one focuses on database integration and element analysis. The architecture combines direct CAD control with SQLite data persistence, enabling complex queries about drawing content and automated element tracking.

Tools available

External APIs and technologies

• AutoCAD COM Interface: Primary control mechanism for AutoCAD automation

• SQLite Database Engine: Local data storage for drawing element analysis

• Python Libraries: pywin32 for COM access, pyautocad for simplified AutoCAD communication, sqlalchemy for database operations, comtypes for additional COM support

Configuration requirements

• Claude Desktop Integration: Requires claude_desktop_config.json modification pointing to executable or Python script

• AutoCAD Requirements: AutoCAD 2018 or higher with COM interface support

• Database Configuration: SQLite database (autocad_data.db) created automatically in server directory

3. autocad-mcp by puran-water (59 ⭐ on Github)

One more AutoCAD server – this one excels at P&ID (Piping and Instrumentation Diagram) creation and process engineering automation. The system translates conversational prompts into AutoLISP code, enabling complex industrial drawing generation with ISA standard compliance.

Tools available

External APIs and technologies

• AutoLISP: Core scripting language for AutoCAD automation

• CAD Tools Online (CTO) P&ID Symbol Library: Commercial library containing 600+ ISA 5.1-2009 standard symbols

• Python Libraries: mcp for protocol implementation, pywin32 for Windows integration, keyboard for input simulation, pyperclip for clipboard operations

Configuration requirements

• CTO Library Installation: Commercial P&ID library at C:\PIDv4-CTO\ for full functionality

• Server Selection:

  • server_lisp_fast.py: Full CTO support with 80% performance improvements

  • server_lisp.py: Basic AutoCAD operations without CTO dependency

• AutoCAD Configuration: Trusted locations setup for LISP file loading, system variables for security and attribute handling

3D parametric modeling platforms

4. autodesk-fusion-mcp-python by sockcymbal (6 ⭐ on Github)

At this stage a bit of a prototype, this cloud-native Fusion 360 integration demonstrates modern CAD automation through Autodesk Platform Services. The server combines local Fusion 360 control with cloud-based design automation, offering both immediate interaction and scalable processing capabilities. However, it only implements one tool at the time of writing:

Tools available

External APIs and technologies

• Autodesk Platform Services (APS): Cloud-based design automation for scalable processing

• Autodesk Fusion 360 API: Direct scripting interface for local Fusion 360 control

• OAuth v2: Secure authentication mechanism for APS access

• Python Libraries: httpx for HTTP client operations, python-dotenv for environment management

• HTTP Servers: Communication bridges between MCP client and Fusion 360 instances

Configuration requirements

• Environment variables:

  • APS_CLIENT_ID: Autodesk Platform Services client identifier

  • APS_CLIENT_SECRET: Associated secret key for authentication

  • FUSION_ACTIVITY_ID: Design automation activity identifier

  • APS_BASE: Optional base URL override for APS endpoints

The server provides three operational modes: cloud automation via APS, local Fusion 360 control through the LiveCube.py add-in, and intermediary bridging via fusion_server.py. This might be a project that is looking for contributors in order to flesh out its capabilities! 

5. fusion360-mcp-server by ArchimedesCrypto (26 ⭐ on Github)

This comprehensive Fusion 360 server provides extensive 3D modeling capabilities through a well-structured API. The system generates Python scripts for complex modeling operations, supporting everything from basic sketching to advanced surface creation.

Tools available

External APIs and technologies

• Autodesk Fusion 360 API: Native scripting interface for model generation

• FastAPI: Modern web framework for creating responsive API endpoints

• Uvicorn: High-performance ASGI server for FastAPI applications

• Pydantic: Data validation and parsing using Python type annotations

Configuration requirements

• Operation Modes:

  • HTTP Server Mode: Direct API access via web endpoints

  • MCP Server Mode: Integration with AI clients using --mcp flag

• Tool Registry: tool_registry.json defines available tools and parameter specifications

• Script Generation: Template-based Python code generation for Fusion 360 operations

Open source CAD integration

6. freecad-mcp by neka-nat (165 ⭐ on Github)

We love to see integrations and support for open-source projects. This FreeCAD integration provides comprehensive control over the popular open-source parametric modeling software. The server enables sophisticated 3D modeling operations while maintaining access to the extensive FreeCAD parts library for rapid prototyping.

Tools Available

External APIs and technologies

• FreeCAD Python API: Core scripting interface for parametric modeling

• SimpleXMLRPCServer: Communication protocol between MCP server and FreeCAD addon

• FastMCP: Framework for building MCP-compliant servers

• GitHub Integration: Direct access to FreeCAD-library repository for parts

Configuration requirements

• FreeCAD Addon Installation: FreeCADMCP addon placed in appropriate Mod/ directory

• RPC Server Activation: Manual startup of RPC server within FreeCAD workbench

• Claude Desktop Configuration: uvx command integration for package management

• Optional Parameters: --only-text-feedback flag for token conservation

7. freecad_mcp by bonninr (68 ⭐ on Github)

This streamlined FreeCAD server focuses on programmatic control and scene analysis. The two-tool approach emphasizes flexibility, allowing direct Python script execution within the FreeCAD environment for maximum customization.

Tools available

External APIs and technologies

• FreeCAD Python API: Direct integration with parametric modeling capabilities

• FastMCP: MCP server framework for protocol compliance

• Python Socket Module: Network communication between server and FreeCAD

• JSON: Data interchange format for commands and responses

Configuration requirements

• Installation: Repository placement in FreeCAD modules directory

• Workbench Activation: Selection of "FreeCAD MCP" workbench after restart

• Path Configuration: Absolute paths required for Python interpreter and script locations

Specialized CAD platforms

8. mcp-server-solidworks by eyfel (14 ⭐ on Github)

This innovative SolidWorks integration takes a contextual approach, acting as an intelligent bridge between AI assistants and the professional CAD environment. Rather than exposing discrete tools, the server provides comprehensive context awareness and action translation.

Core capabilities

External APIs and technologies

• SolidWorks API: Professional CAD control through comprehensive object model

• COM Interface: Windows component architecture for SolidWorks communication

• PythonNET: Bridge technology connecting Python logic with .NET SolidWorks adapters

• C# Adapter Libraries: Version-specific DLLs for SolidWorks compatibility

Configuration requirements

• Version-Specific Adapters: Separate DLLs for SolidWorks 2021, 2022, 2023, etc.

• Configuration Directory: Settings for SolidWorks version, communication ports, and model identifiers

• Claude Model Integration: API credentials for LLM communication

This implementation stands out in the uniqueness of its approach to integrating with the underlying software. 

9. onshape-mcp by BLamy (3 ⭐ on Github)

Onshape is cloud-native CAD software. This MCP server integrates with Onshape to provide extensive control over the browser-based CAD platform. The server delivers comprehensive document management, feature manipulation, and assembly operations through Onshape's robust REST API.

Tools available

External APIs and technologies

• Onshape REST API: Comprehensive cloud CAD control through HTTP endpoints

• Node.js: JavaScript runtime for server-side operations

• TypeScript: Enhanced JavaScript with static typing for robust development

• Zod: Schema validation library ensuring data integrity

• node-fetch: HTTP client for Onshape API communication

Configuration requirements

• Environment variables:

  • ONSHAPE_ACCESS_KEY: Personal access key for Onshape API authentication

  • ONSHAPE_SECRET_KEY: Associated secret key for secure communication

  • ONSHAPE_API_URL: Base URL for Onshape API endpoints (default: https://cad.onshape.com/api)

Comparative analysis

Platform coverage and tool density

Technology stack and requirements

MCP ecosystem integration support

Orchestrating CAD workflows

The real power of these MCP servers emerges when multiple tools work in concert, creating engineering workflows that span different CAD platforms and specializations.

Multi-platform design pipeline

Scenario: A mechanical engineer developing a complex assembly from concept through manufacturing.

• Begin with freecad-mcp for initial concept modeling and design exploration using open-source flexibility

• Transfer key components to fusion360-mcp-server for detailed parametric modeling and simulation preparation

• Export manufacturing drawings through autocad-mcp for shop floor documentation and P&ID generation

• Use onshape-mcp for collaborative review and version control with distributed teams

This pipeline leverages each platform's strengths while maintaining design intent across software boundaries.

Process engineering integration

Scenario: Chemical plant design requiring both equipment modeling and process documentation.

• Deploy autocad-mcp with CTO library integration for ISA-compliant P&ID creation

• Model individual equipment using fusion360-mcp-server for detailed component design

• Generate assembly drawings and installation details through solidworks-mcp integration

• Maintain collaborative design reviews using onshape-mcp for stakeholder access

This workflow addresses the complete spectrum from process design to equipment specification.

Automated design variation

Scenario: Product family development requiring multiple design iterations.

• Establish base parametric models using freecad-mcp for foundational geometry

• Generate manufacturing variations through fusion360-mcp-server with different material and size parameters

• Automatically produce technical drawings via autocad-mcp for each design variant

• Coordinate design reviews through onshape-mcp cloud collaboration

This approach enables rapid iteration while maintaining design consistency across product families.

How to make a good MCP server for Computer-Aided Drafting?

MCP servers lower the barrier to CAD proficiency by abstracting complex command structures. Junior engineers can achieve sophisticated modeling results without years of software-specific training. AI assistants are becoming knowledgeable mentors, guiding users through best practices while handling routine operations automatically. After examining these nine MCP servers, several architectural patterns emerge that inform best practices for CAD integration:

Communication Architecture

COM interface pattern: Windows-based CAD applications (AutoCAD, SolidWorks) rely heavily on COM automation for external control. The pywin32 library provides reliable access, but requires careful error handling and resource management to prevent application instability.

REST API pattern: Cloud-native platforms (Onshape, Autodesk Platform Services) offer more robust integration through HTTP APIs. These interfaces provide better error handling and don't require platform-specific dependencies, but introduce network latency considerations.

RPC pattern: Applications like FreeCAD benefit from XML-RPC or similar protocols that provide structured communication without COM complexity. This approach offers good performance for local installations while maintaining cross-platform compatibility.

Script generation strategies

Template-based approach: The most successful servers use template-driven script generation, where predefined code structures are populated with user parameters. This ensures reliable execution while maintaining flexibility for complex operations.

Direct API mapping: Simple servers map MCP tools directly to single API calls, providing reliable operation but limited capability for complex workflows. This approach works well for basic operations but struggles with multi-step processes.

Hybrid command processing: Advanced servers combine template generation with direct API access, enabling both simple operations and complex scripted workflows. This provides maximum flexibility at the cost of implementation complexity.

The future of MCP and computer-aided drafting

The CAD MCP ecosystem demonstrates remarkable potential for transforming engineering workflows. Current limitations around platform coverage and tool sophistication will likely diminish as the technology matures and adoption increases.

Cloud-native CAD platforms show particular promise for MCP integration due to their API-first architectures and reduced dependency on platform-specific technologies. The success of Onshape and Autodesk Platform Services integrations suggests that future CAD tools will prioritize programmatic access alongside traditional user interfaces.

The emergence of specialized servers for process engineering (P&ID tools) and contextual assistants (SolidWorks integration) indicates growing sophistication in domain-specific applications. We can expect similar specialization for other engineering disciplines like electrical design, architecture, and simulation.

As these MCP servers evolve, they promise to make sophisticated CAD capabilities accessible to a broader range of professionals while enabling veteran engineers to focus on creative problem-solving rather than repetitive software manipulation. The combination of AI understanding with precise engineering tools opens new possibilities

How Snyk can help you secure your AI-generated 3D modeling workflows

AI tools bring exciting possibilities to 3D modeling, but security considerations remain important, especially when implementing custom code in the scripts called by your MCP server; or sharing models online on platforms like Onshape. Snyk can scan for vulnerabilities in your projects, protecting both your development environment and your creative assets.

For enterprise-grade security without the enterprise-level costs, open source projects can apply for free access to Snyk's premium tools through the Secure Developer project. Check out some of the projects that have already joined the initiative!