SidequestLab

FDTD Lab MCP

In Progress

FDTD Lab MCP is an MCP server that enables AI agents to safely operate Ansys Lumerical FDTD workflows. It connects .fsp inspection, object/property access, protected run-directory creation, parameter sweeps, simulation execution, monitor result retrieval, and CSV export into one automation flow. The project is currently stabilizing real-adapter startup and timeout behavior while maintaining a fake adapter for development and CI.

Key Features

.fsp inspection

Open FDTD projects and summarize object, region, and monitor structure

Property automation

Expose object property reads and edits as MCP tools

Protected run directories

Preserve source files while creating experiment-specific run folders

Parameter sweeps

Propose, validate, run, and collect sweep results

Adapter split

Separate fake, ansys-lumerical-core, and lumapi fallback paths for development and real environments

Reliability work

Stabilizing real-adapter startup latency and timeout handling

Tech Stack

PythonMCPAnsys Lumerical FDTDansys-lumerical-corelumapipytest