Data Cambridge

Several tools overlap heavily in purpose: ask_pipeworx, ask_pipeworx_beta, ask_pipeworx_grounded, and deep_research all route questions to the same 5,961-source backend, and the beta description admits it currently matches the stable router exactly. Discovery tools (discover_tools vs suggest_questions) and entity tools (entity_profile vs compare_entities vs recent_changes vs resolve_entity) also have fuzzy boundaries. Descriptions work hard to differentiate, but an agent can easily pick a suboptimal tool.

Nearly all tools use snake_case, with clear domain prefixes like polymarket_, cambridge_, ask_pipeworx, and scan_. However, verb usage is inconsistent — some are verb_noun (list_subscriptions, validate_claim, compare_entities), while others are bare nouns (entity_profile, deep_research, recent_changes, search_within) or single verbs (remember, forget, subscribe). The style is readable and mostly predictable, just not a single uniform pattern.

34 tools is very heavy for an agent to navigate, and several appear redundant (e.g., ask_pipeworx vs ask_pipeworx_beta, multiple overlapping discovery and entity tools). While the server spans many domains (Cambridge open data, Pipeworx routing, Polymarket, memory, subscriptions), the surface feels bloated rather than well-scoped. A leaner set of meta-tools would reduce selection cost without losing capability.

The surface covers a broad data-gateway lifecycle: discovery (suggest_questions, discover_tools), querying (ask_pipeworx, deep_research, cambridge_query), entity resolution/profiling (resolve_entity, entity_profile, compare_entities), fact-checking (validate_claim), memory CRUD (remember/recall/forget), and subscription lifecycle (subscribe/list_subscriptions/recent_alerts/unsubscribe). Minor gaps exist, such as no direct tool to invoke arbitrary discovered Pipeworx tools (agents must route through ask_pipeworx), but workarounds are available.