Hyperbrowser

Most tools have clearly distinct purposes, and the long descriptions do a good job of steering agents (e.g., ask_pipeworx vs ask_pipeworx_grounded vs deep_research). However, there is meaningful overlap among the prediction-market tools (polymarket_edges, polymarket_arbitrage, polymarket_edge_tracker, bet_research) and between discovery tools (discover_tools vs suggest_questions), so misselection is still plausible.

Names are uniformly lowercase snake_case, but conventions are otherwise mixed: some tools are prefixed by domain (hyperbrowser_*, polymarket_*, pipeworx_*, resolution_*), some are bare verb-noun phrases (compare_entities, resolve_entity, validate_claim), and some are noun-first compounds (company_facts, bet_research, kalshi_weather_edge). The set is readable, but there is no consistent verb_noun pattern across the whole surface.

39 tools is well above the 25+ threshold where a toolset starts to feel bloated, especially when many are meta-tools (ask_pipeworx variants, discovery, feedback, trending) layered on top of already-specific data tools. The count is justified by the broad data domain, but the exposed surface would be easier to navigate if consolidated.

The surface is highly comprehensive for its stated domain: query, research, comparison, entity resolution, memory, subscriptions, feedback, and browser actions are all covered with few dead ends. Minor gaps exist (e.g., no browser automation actions beyond scrape/crawl/extract, no direct trading execution), but agents can accomplish the core workflows without workarounds.