Lob

Several tool families overlap heavily: ask_pipeworx / ask_pipeworx_beta / ask_pipeworx_grounded / deep_research / validate_claim all answer questions from routed sources, and polymarket_edges / polymarket_arbitrage / polymarket_fill_risk / polymarket_edge_tracker / polymarket_kalshi_spread all scan related prediction-market edges. The descriptions do provide preference guidance, but an agent still has to read carefully to choose correctly.

Nearly all tools use consistent snake_case, with predictable patterns like lob_verify_address, resolve_entity, polymarket_fill_risk, and recent_changes. Minor deviations such as ask_pipeworx_beta, generate_llms_txt, and scan_dependency are readable and do not break the convention.

39 tools is heavy for a single MCP server, especially because the set mixes Lob address verification, a broad Pipeworx data gateway, memory, subscriptions, AI visibility, and multiple prediction-market analytics suites. The count increases selection burden and exceeds the typical well-scoped 3–15 range.

For a universal data/research gateway the surface is broad: discovery, question routing, grounded answers, deep research, entity resolution, financial facts, prediction-market analysis, memory, subscriptions, and feedback are all covered. Some areas are explicitly incomplete or empty, such as external resource creation and the court release category, but agents have workarounds via meta-tools.