MCP server¶
lightsmith-mcp is a Model Context Protocol server. It lets AI agents
build, check, solve and compare Lightsmith problems through the same JSON interface as the Python API and
the CLI.
To use it, register the lightsmith-mcp command as a stdio server in your MCP client.
What the server offers¶
The MCP server is being expanded, so this page describes it only in general terms. The full reference will be on its own page when that work is merged. In this release the server covers:
- Discovery: the JSON Schema of every problem kind, the installed backends, and the CPU and GPU environment.
- Pre-flight: estimates of size, device and memory, without solving.
- Solving: one method, or several methods with relative differences against a reference.
Results are the same JSON models the Python API returns, including provenance and warnings.
Recommended agent workflow¶
- Read the schema before building a problem.
- Use nm and vacuum wavelengths, with
Im(ε) > 0for loss (see Conventions). - Estimate before solving grid-based problems.
- Solve, then check
warningsbefore reporting numbers. - Cross-check important results with a second method.
The repository's AGENTS.md and llms.txt give the same guidance in compact form.