"author:Presciman" matching MCP servers:

Enables LLM agents to answer questions about Darshan I/O logs by first opening a log and then retrieving exact, tool-computed numbers: canonical per-concern cards (volume, bandwidth, request sizes, metadata, rank balance, Lustre, DXT, verdict), an outline/search/get interface over a two-layer schema, and exact arithmetic, so answers are retrieved rather than estimated.