Agentic Phased Array Builder
- Python
- LLM agents
- MCP
- EdgeFEM
- Phased-array modeling
Problem
Antenna design spans several tools and a lot of manual coordination: design a unit cell, build the array, then roll the results up into system-level metrics. Each step lives in a different model, and moving between them by hand is slow and easy to get wrong.
Approach
This project wires that pipeline behind an LLM agent. The agent exposes the design tools over the Model Context Protocol and runs the chain itself, from unit cell to array to system metrics, calling an electromagnetic solver (EdgeFEM) and a phased-array model as tools. A request in plain language becomes a sequence of solver runs and a design rolled up to the numbers an engineer actually wants.
Result
It connects two things I care about: agentic AI and antenna physics. The agent handles the tool-to-tool plumbing that used to take manual effort, while the underlying solvers keep the results grounded in electromagnetics rather than a language model’s guesswork. The repo shows the agent architecture and the MCP tool definitions.