MathWorks engineers demonstrate how to use Claude Code with the MATLAB MCP Server to build Simulink and Simscape models through an agentic AI workflow. Using a coffee-cup cooling simulation as a controlled example, the experiment shows how AI agents can construct models incrementally via MATLAB APIs while preserving engineering rigor, traceability, and validation. A skills.md file encodes modeling conventions so the agent follows established practices rather than improvising. The approach positions AI as an accelerator for repetitive construction tasks while engineers retain control over structure and verification.
Table of contents
Experiment overviewEnvironment configurationPlanning the modelDefining skills for grounded modelingIncremental development and validationFinal modelWhat this demonstratesWhy this matters84 Impressions