Plaid

Several tools can answer the same kind of question: ask_pipeworx, ask_pipeworx_beta, ask_pipeworx_grounded, deep_research, and validate_claim all route or verify factual queries, and the beta variant currently matches ask_pipeworx exactly. The prediction-market cluster (polymarket_edges, polymarket_arbitrage, polymarket_fill_risk, polymarket_kalshi_spread, bet_research) and company-data cluster (entity_profile, company_facts, compare_entities, recent_changes) also require careful reading to disambiguate.

All names are lowercase snake_case and several families share clear prefixes (ask_pipeworx*, polymarket_*, pipeworx_*, resolution_*), which helps. However, the convention is mixed: verb-first names like get_accounts, compare_entities, and validate_claim sit alongside noun-phrase names like entity_profile, company_facts, and kalshi_weather_edge, and the memory trio uses remember/recall/forget rather than a consistent set/get/delete style.

39 tools is well past the 25+ threshold where a server starts feeling heavy, and many of them are variants or wrappers—three ask_pipeworx routers and six overlapping Polymarket tools could be consolidated. The broad data-domain scope explains some of the count, but a server named Plaid only has three actual Plaid operations, making the overall surface feel overbuilt and unfocused.

The toolkit covers a lot of ground: grounded Q&A, company profiles, entity resolution, prediction-market edge analysis, subscription lifecycle, and memory CRUD are all present. However, the Plaid side is limited to accounts, balances, and transactions with no item/link lifecycle or broader Plaid products, and the pipeworx:// citation URIs returned by the data tools have no obvious fetch-by-URI tool to resolve them—leaving notable dead ends.