🔗 SimuBridge

AI-Powered Simulink Skill Pack with MCP Backend

中文文档

What is this

simubridge-skills is a Claude Code / Codex skill project that lets AI assistants control MATLAB Simulink using natural language. It contains:

• 🎯 Skill pack (skill/simubridge/) — prompt bundle that teaches AI how to use Simulink tools effectively

• 🔧 MCP server (src/simubridge/) — local backend connecting to MATLAB Engine with 20+ tools

The skill is the product. The MCP server is the backend.

Related MCP server: simulink-mcp

Skill Index

Quick Start

Prerequisites

• MATLAB R2021a+ with Simulink

• Python 3.9–3.12

1. Install MATLAB Engine for Python

cd "C:\Program Files\MATLAB\R2024a\extern\engines\python"
python setup.py install

Verify: python -c "import matlab.engine; print('OK')"

💡 Permission error? Add --user. Python 3.13+? Downgrade to 3.11 or 3.12.

2. Install the MCP Backend

git clone https://github.com/naaomiur/simubridge-skills.git
cd simubridge-skills
pip install -e .

3. Configure MATLAB

Run in MATLAB:

matlab.engine.shareEngine('SIMULINK_MCP_SESSION')

💡 Add to startup.m for auto-share on every launch:

edit(fullfile(userpath, 'startup.m'))

4. Install the Skill

The skill teaches Claude how to use Simulink tools effectively. Copy the entire folder (not just SKILL.md) — the references/ directory is needed by the skill.

⚠️ Only copying SKILL.md will silently break the skill. Always copy the whole skill/simubridge/ folder.

Claude Code (with MCP — full functionality)

mkdir -p ~/.claude/skills
cp -R skill/simubridge ~/.claude/skills/

Configure the MCP server (see Step 5), restart, and Claude can fully control Simulink.

Claude Code (standalone skill — no MATLAB needed)

Even without MATLAB or the MCP backend, you can still use the skill as a Simulink knowledge base for planning and discussion:

mkdir -p ~/.claude/skills
cp -R skill/simubridge ~/.claude/skills/

Start a Claude Code session and ask:

Read the simubridge skill. I'm designing a motor control system in Simulink —
suggest a block diagram layout with the right blocks and wiring topology.

Claude will read SKILL.md and references/tool-guide.md to give informed guidance — even without executing anything. This is useful for:

• Planning model architecture before building it

• Getting block library paths and parameter names for manual use

• Learning Simulink workflows and best practices

• Designing subsystem hierarchies and wiring strategies

💡 How it works: The skill provides Claude with detailed knowledge of every Simulink tool, common library paths, port types, and wiring patterns. Claude can then advise you on model design even when the MCP backend isn't running.

Codex

⚠️ Codex is not yet supported. We are working on Codex skill packaging. For now, use Claude Code.

5. Configure the MCP Server

The MCP server must be configured so the AI assistant can connect to it.

Claude Code

Add to claude_desktop_config.json:

• Windows: %APPDATA%\Claude\claude_desktop_config.json

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "simubridge": {
      "command": "python",
      "args": ["-m", "simubridge"]
    }
  }
}

You can also copy the included claude_desktop_config.json directly.

Codex

Add the same configuration in Codex's MCP settings panel.

Verify Installation

Restart your AI assistant. Start a new session and try:

Open mymodel.slx, add a Gain block set to 2.5 after the Voltage Source,
and connect it to a Scope

If everything is configured correctly, Claude will:

1. Recognize the Simulink task

2. Read the skill definitions from SKILL.md

3. Call MCP tools like search_library, add_block, connect to complete the task

What the Skill Can Do

📖 Complete tool documentation: skill/simubridge/SKILL.md

Architecture

┌─────────────────┐     Skill      ┌──────────────┐     MCP stdio     ┌──────────────┐   MATLAB API   ┌──────────┐
│  Claude / Codex  │ ◄────────────  │  simubridge   │ ◄──────────────► │  MCP Server   │ ◄────────────► │  MATLAB  │
│  + Skill Prompt  │                │  (skill def)  │                  │  (Python)     │                │(Simulink)│
└─────────────────┘                └──────────────┘                  └──────────────┘                └──────────┘

• Skill layer: skill/simubridge/SKILL.md — teaches AI how to use each tool

• Transport: MCP stdio — local process, no network

• Engine: matlab.engine API — connects to your running MATLAB session

• Auto-save: every write operation auto-saves the model

Project Structure

simubridge-skills/
├── skill/simubridge/               # 🎯 Skill pack (primary)
│   ├── README.md                    #    Skill documentation
│   ├── SKILL.md                     #    Skill definition (bilingual tool reference)
│   └── references/
│       └── tool-guide.md            #    Quick reference
├── src/simubridge/                  # 🔧 MCP backend
│   ├── __init__.py                  #    Entry point + tool registration
│   ├── app.py                       #    FastMCP + MATLAB engine manager
│   └── tools/                       #    7 tool modules
├── examples/
│   └── basic_usage.md
├── pyproject.toml
├── claude_desktop_config.json
├── LICENSE
├── README.md                        #    This file (English)
└── README_CN.md                     #    Chinese version

License

MIT License — see LICENSE.

⚠️ MATLAB and Simulink are registered trademarks of The MathWorks, Inc. This project is not affiliated with or endorsed by MathWorks.

Acknowledgments

Built with FastMCP and the Model Context Protocol.