Pypi Stats

Most tools have distinct purposes, but the research-routing cluster (ask_pipeworx, ask_pipeworx_beta, ask_pipeworx_grounded, deep_research) overlaps heavily — the beta tool is documented as currently identical to the stable router, which is a genuine misselection trap, and the five polymarket_* tools plus discover_tools/suggest_questions sit close together. The verbose descriptions do a good job of spelling out when to prefer each one, which keeps this from dropping lower.

The dominant convention is snake_case verb_noun (resolve_entity, compare_entities, scan_dependency, list_subscriptions), applied consistently across most of the surface. However, several tools break the pattern with bare nouns or no verb at all (overall, recent, system, entity_profile, recent_alerts), and the ask_pipeworx family relies on suffix variation rather than a clear verb. Readable throughout, but not a predictable pattern.

36 tools is well above the comfortable range for a single server, and the name 'Pypi Stats' accounts for only five of them. The remainder is a broad data-gateway surface (entity resolution, comparison, prediction markets, memory, subscriptions) where each tool does have a rationale, so it is heavy rather than padded — but the count is still awkward for an agent to reason over.

The stated PyPI stats domain is fully covered (overall, recent, python_major, python_minor, system mirror the pypistats.org endpoints), and the surrounding gateway surface is broadly complete: memory has save/read/delete, subscriptions have create/list/cancel plus an alerts read path, and entity work spans resolve → profile → compare → changes → validate. Minor gaps remain (no direct raw-record fetch primitive beyond search_within, no PyPI package-metadata tool), but nothing that creates a hard dead end.