This toolkit empowers AI coding agents with the tools and expertise to effectively interact with MATLAB Simulink models and adhere to Model-Based Design (MBD) best practices. It enables agents to read, build, edit, and test complex Simulink models, significantly enhancing automation capabilities for engineers and researchers.

How It Works

The toolkit connects AI agents (e.g., Claude, Copilot, Gemini) to Simulink via the Model Context Protocol (MCP). A MATLAB-hosted MCP Server acts as the bridge, translating agent requests into actions within Simulink. Agents utilize 6 MCP tools for direct model manipulation and are guided by 7 agent skills that encapsulate MBD expertise, facilitating sophisticated MBD workflows.

Quick Start & Requirements

Automated setup is recommended via the setupAgenticToolkit("install") MATLAB function, which installs the MCP server and configures the agent. Prerequisites include MATLAB R2023a or later with Simulink, and a supported AI coding agent. Optional dependencies include Simulink Test, System Composer, Simscape, and Stateflow. Detailed instructions are available in the Getting Started guide.

Highlighted Details

• MCP Tools: Offers model_overview, model_read, model_edit, model_test, model_query_params, and model_resolve_params for comprehensive model interaction.
• Agent Skills: Encompasses MBD best practices for model building, simulation, specification, testing, and bug reporting.
• AI Agent Support: Verified with Claude Code, GitHub Copilot, OpenAI Codex, Gemini CLI, and Sourcegraph Amp.
• Agentic Task Explorer: A research preview showcasing multi-step agent tasks for Simulink workflows.
• Model Capability: Higher-capability AI models are recommended for optimal performance on complex tasks.

Maintenance & Community

Feedback is welcomed via GitHub Issues. While pull requests are reviewed, external contributions are not merged. No specific community channels (like Discord/Slack) or roadmaps are detailed in the README.

Licensing & Compatibility

MCP servers are licensed under the MathWorks Software License Agreement and are strictly permitted for use only with MATLAB and Simulink. Shared or centralized server use requires explicit contact with MathWorks.

Limitations & Caveats

The toolkit is under active development, and the Agentic Task Explorer is a research preview. Performance is highly dependent on the AI model's reasoning and coding capabilities; lower-capability models may yield less reliable results. Security requires careful validation of all tool calls, with a human-in-the-loop recommended for critical actions.