Committee Releases

Many tools are distinct (remember/recall/forget, subscribe/unsubscribe/list_subscriptions, resolve_entity), but the routing layer overlaps badly: ask_pipeworx, ask_pipeworx_beta (explicitly 'currently matches ask_pipeworx exactly'), ask_pipeworx_grounded, and deep_research all answer questions from the same catalog, and discover_tools vs suggest_questions both serve discovery. The six polymarket_* tools plus bet_research also blur together for prediction-market queries, forcing agents to parse long descriptions to choose.

Nearly everything is consistent snake_case, mostly verb_noun (resolve_entity, validate_claim, compare_entities, search_within, scan_dependency) or clear namespaced patterns (polymarket_*, subscribe/unsubscribe). Minor deviations are noun-phrase names like recent_alerts, recent_changes, entity_profile, and ai_visibility_check, but the overall convention is predictable.

34 tools is heavy, and they span wildly unrelated domains — congressional committee documents, prediction-market arbitrage, LLM visibility probing, SEC/EDGAR profiling, memory, and subscription management. For a server literally named 'Committee Releases' only three tools (search_committee_documents, get_committee_document, list_committee_document_types) fit the stated purpose, so most of the surface feels tacked on.

For its stated committee-document niche the coverage is decent (search, read, list types, with citation guidance), but it explicitly does not extract PDF text, which is where most committee letters and transcripts live — a real dead end for the core domain. The broader gateway surface is broad but uneven, with soft-failing sources (USPTO, PatentsView sunset) noted throughout.