Ferc Eqr

Several clusters have fuzzy boundaries: ask_pipeworx, ask_pipeworx_beta, ask_pipeworx_grounded, and deep_research are near-siblings where the beta router is essentially identical to the stable one, and entity_profile / company_facts / compare_entities / recent_changes / validate_claim all touch company research. The descriptions work hard to steer selection, so it is not chaotic, but an agent must read long prose to avoid misselecting the wrong router or company tool.

All names are snake_case, which is consistent, but the verb/noun conventions are mixed: verb_noun (eqr_contracts_search, scan_dependency, resolve_entity) alongside bare noun phrases (entity_profile, company_facts, recent_changes, ai_visibility_check). Domain prefixes (eqr_, polymarket_, ask_pipeworx*) are used unevenly, so the pattern is readable but not predictable.

40 tools is heavy even for a broad gateway, well above the 25-tool threshold for 'too many.' Breadth is the point of the server, but the count forces review overhead and amplifies the overlapping-router problem, and several meta-tools exist mainly to help agents cope with the volume.

Coverage is broad and lifecycle-complete where it matters: memory (remember/recall/forget), subscriptions (subscribe/list_subscriptions/unsubscribe/recent_alerts), discovery, company/financial, prediction-market, and research surfaces all have create/read/delete paths. The main gaps are acknowledged — FERC EQR is Phase 1 only (contract terms, no transaction price/volume) and some upstream sources soft-fail — but no domain has an obvious dead end.