Onet

Multiple tool clusters overlap heavily: ask_pipeworx, ask_pipeworx_beta, ask_pipeworx_grounded, and deep_research all route queries to the same underlying data, and the prediction-market tools (polymarket_arbitrage, polymarket_edges, polymarket_edge_tracker, polymarket_fill_risk, polymarket_kalshi_spread) have fuzzy boundaries despite detailed descriptions. Even with rich docs, an agent could easily select the wrong variant for a given task.

All tool names are snake_case, but they mix action verbs (ask_, scan_, validate_, resolve_) with noun-phrase names (company_facts, entity_profile, kalshi_weather_edge, release_calendar_markets). The polymarket_* and onet_* prefixes create sub-families, but overall there is no single consistent verb_noun or noun_verb pattern across the set, making it somewhat predictable but not fully uniform.

40 tools is far beyond the 3-15 well-scoped range and even exceeds the 25-tool 'too many' threshold. The server tries to cover AI visibility, structured data querying, prediction markets, O*NET occupations, memory, subscriptions, and feedback—sprawling across multiple unrelated domains, which feels like a kitchen sink rather than a focused toolset.

The server provides broad coverage within its sub-domains: query routing, company profiles, prediction-market analysis, O*NET lookups, memory, and subscriptions all have associated operations. However, the overall domain is undefined and ad-hoc, and some areas feel incomplete (e.g., only four O*NET tools, no direct market-order placement, no tool to cancel subscriptions except unsubscribe). This is more of an eclectic bundle than a coherent lifecycle.