Skip to content

lumerical-mcp v0.4.0

Choose a tag to compare

@davydov-nikita davydov-nikita released this 04 Jul 12:57
· 3 commits to main since this release

Initial public release.

An MCP server that lets an AI assistant drive ANSYS Lumerical FDTD: build scenes, run simulations, and read results back — with a correctness layer that catches settings which silently produce wrong physics.

Highlights

  • 87 async tools over the official Python lumapi: geometry, sources, monitors, mesh and boundaries, ports/S-parameters, parametric sweeps, field-to-PNG rendering, run progress/cancel.
  • Materials including sampled (n,k) tables and full anisotropy: arbitrary 3×3 hermitian permittivity tensors and magneto-optic (gyrotropic) media via the grid-attribute matrix transform.
  • Correctness layer: validate_setup (pre-flight physics checks with fixes), assess_results (post-run verdict incl. energy balance), diagnose (error → cause + fix), convergence_test/certify (trust verdict for any figure of merit).
  • Figure-of-merit extractors: Q-factor, Purcell factor, scattering/absorption cross-sections, grating orders, far-field, photonic bands, χ² second harmonics.
  • Benchmark suite: Lumerical Application Gallery examples rebuilt through the tools from written specs and scored against the gallery's own references.
  • 200+ mocked unit tests run without a Lumerical install; a live physics panel gates releases on machines with a license.

Install

pip install https://github.com/davydov-nikita/lumerical-mcp/releases/download/v0.4.0/lumerical_mcp-0.4.0-py3-none-any.whl

or from source: pip install git+https://github.com/davydov-nikita/lumerical-mcp

Requires a local ANSYS Lumerical FDTD install (developed against 2024 R1 / v241) and an MCP client. See the README for client configuration.