"author:okabrionz" matching MCP servers:

Lets agents inspect a stored crawl through twelve tools: crawl and issue summaries, tab-and-filter URL queries, per-page measurements, root-cause clusters, prioritised repair plans with generated markup, live re-checks of a single URL, diffs between two crawls, bounded live crawls, and report exports. This grounds an agent's answers in measured crawl data rather than recalled guesses, while clearly labelling what is inferred.