CATIA Mcp Ai Automation

🚀 CATIClaude

AI-Powered CATIA V5 Automation through the Model Context Protocol

⭐ Badges

🧠 Overview

CATIClaude is a production-grade Model Context Protocol (MCP) server that connects Claude Desktop directly to CATIA V5 through pyCATIA COM automation.

It enables engineers to design real mechanical parts using natural language — and automatically generate:

CATPart files STEP / IGES exports Fully parametric 3D geometry

Example:

“Design a lightweight L-bracket with 120mm base, 10mm thickness, and 2 mounting holes”

➡️ CATIA automatically builds the full 3D model.

🏛 Design Philosophy

CATIClaude is designed around one central principle:

> AI should describe engineering intent—not low-level CAD operations.

Instead of exposing the complexity of the CATIA COM API to the user, CATIClaude translates natural language into deterministic CAD operations through a layered planning architecture.

This keeps the interface simple while maintaining engineering reliability.

⚙️ Key Idea

Instead of manually modeling in CAD software:

🧠 English → AI Planning → MCP Tools → CATIA V5 → Real Geometry

🚀 Roadmap

• [x] CATIA V5 integration
• [x] MCP Server
• [x] Natural-language CAD
• [x] STEP export
• [x] Parametric editing
• [ ] Assembly generation
• [ ] Drafting automation
• [ ] Multi-agent engineering workflow
• [ ] AI optimization tools
• [ ] CATIA V6 support

🏗️ System Architecture

flowchart TD
    A[Claude Desktop] --> B[MCP Server Core]
    B --> C[CAD Intent Planner]
    C --> D[Execution Engine]
    D --> E[pyCATIA Abstraction Layer]
    E --> F[CATIA V5 COM API]
    F --> G[3D Model + Export Files]

    style A fill:#ff7a18,stroke:#333,color:#fff
    style G fill:#1f77ff,stroke:#333,color:#fff

🧩 Architecture Layers

| Layer | Component | Responsibility | | ----- | -------------------- | ------------------------------------ | | 1 | MCP Server Core | Tool routing + FastMCP interface | | 2 | CATIA Session Engine | Stable COM session management | | 3 | pyCATIA Abstraction | Safe wrapper over CATIA API | | 4 | CAD Intent Planner | Natural language → structured design | | 5 | Execution Engine | Step-by-step CAD construction |

📁 Project Structure

catia_mcp/

├── catia/ ├── core/ ├── engine/ ├── tools/ ├── utils/ ├── tests/ ├── config/ ├── server.py ├── requirements.txt └── README.md

🧠 Core Features

🔹 Natural Language

CAD Convert text → real CAD geometry Supports brackets, plates, shafts, housings

🔹 Robust CATIA Session Engine

Singleton COM connection Auto-reconnect on crash Safe multi-call execution

🔹 Sketch Intelligence

Automatic sketch closure validation Prevents invalid Pad/Pocket operations

🔹 Parametric Design Support

Named parameters Real-time geometry updates

🔹 Export Pipeline

STEP / IGES / STL export Automatic file generation

✨ Features

• Natural-language CAD generation
• Rule-based CAD planning engine
• Stable CATIA COM session management
• Automatic geometry validation
• Parametric modeling support
• STEP / STL / IGES export
• Thread-safe MCP execution
• Retry & recovery system
• Structured JSON responses
• Mock-based testing
• Production-ready architecture

🔄 Execution Pipeline

User Prompt

↓

Intent Parser

↓

CAD Planner

↓

Validator

↓

Executor

↓ pyCATIA Wrapper

↓

CATIA V5 Model

↓

Exported CAD Files

🛠 Available MCP Tools

Document

• create_part
• open_document
• save_document
• close_document

Sketch

• create_sketch
• add_line
• add_circle
• add_rectangle
• close_sketch

Features

• pad
• pocket
• hole
• fillet
• chamfer

Parameters

• set_parameter
• get_parameter
• list_parameters

Analysis

• get_tree
• get_edges
• measure_distance
• getmassproperties
• validate_geometry

Export

• export_step
• export_iges
• export_stl

AI Design

• designfromtext

💡 Example Usage

1. L-Bracket

Design a lightweight L-bracket with: Base length: 120mm Thickness: 10mm 2 mounting holes

1. Shaft

Create a 20mm diameter, 150mm long steel shaft.

1. Enclosure

Design a hollow box: 150 × 90 × 60 mm Wall thickness: 3mm Export STEP file

🧪 Reliability Engineering

🔁 Automatic retry system (COM-safe)

🧯 Crash recovery for CATIA sessions

🧩 Geometry validation before operations

🧠 Structured error system (no raw tracebacks)

🔒 Global execution lock (thread-safe MCP)

📜 Full audit logging for every CAD action

⚠️ CATIA Compatibility Note

Some CATIA V5 versions differ in COM APIs.

To ensure stability:

Hole feature is implemented via sketch + pocket fallback Fillet/Chamfer uses stable ShapeFactory calls Only feature_manager.py may require version-specific tweaks

🧰 Installation

Requirements

Windows 10,11 / CATIA V5 (R20+ recommended) / Python 3.10+ / pywin32 / pycatia

Setup

git clone https://github.com/YOUR_USERNAME/catia-claude-mcp cd catia-claude-mcp

pip install -r requirements.txt python -m pywin32_postinstall -install

Run Server

python server.py

Connect to Claude Desktop

Edit:

%APPDATA%\Claude\claudedesktopconfig.json

JSON :

{ "mcpServers": { "catia": { "command": "C:\\Python\\python.exe", "args": ["C:\\path\\to\\catia_mcp\\server.py"] } } }

📦 Output

.CATPart

.STEP

.IGES

.STL

🧪 Testing

pytest -q

Mock-based tests (no CATIA required).

🌍 Why This Project Matters

This project explores:

AI × Mechanical Engineering automation

Natural language CAD systems

Multi-agent engineering workflows

Future of design engineering interfaces

🤝 Contributing

Contributions are welcome.

If you would like to improve CATIClaude:

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Open a Pull Request

📜 License

This project is released under the MIT License.

See the LICENSE file for details.

🙏 Credits

Built with:

• Claude Desktop
• Model Context Protocol (MCP)
• pyCATIA
• Python
• CATIA V5 COM Automation

Special thanks to the open-source communities behind these projects.

📖 Citation

If you use this project in your research or engineering workflow, please consider citing the repository and giving it a GitHub star.

👤 Author

Pouya Bagheri

Mechanical Engineering Student

Interests

• AI for Engineering
• CAD Automation
• Mechanical Design
• Digital Manufacturing
• Model Context Protocol
• Engineering Intelligence

LinkedIn: www.linkedin.com/in/pouyabagheri

Gmail: pouya.bagheri.2005@gmail.com

---

If this project helped you, consider giving it a ⭐ on GitHub.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: pouyabagheri2005-sudo
• Source: pouyabagheri2005-sudo/CATIA-mcp-ai-automation
• License: MIT
• Homepage: https://github.com/pouyabagheri2005-sudo/CATIClaude

Install and usage instructions live in the source repository linked above.