Zyte

Several clusters of tools have unclear boundaries: ask_pipeworx, ask_pipeworx_beta (currently identical), ask_pipeworx_grounded, and deep_research all route questions to the same underlying catalog; discover_tools and suggest_questions both serve tool-discovery; and there are seven-plus prediction-market tools (polymarket_edges, polymarket_arbitrage, polymarket_edge_tracker, polymarket_fill_risk, polymarket_kalshi_spread, bet_research, kalshi_weather_edge) whose scopes overlap heavily despite detailed descriptions. An agent would frequently need to read long descriptions to pick correctly, and some choices are genuinely ambiguous.

Tool names overwhelmingly follow snake_case with a clear verb_noun or prefix_noun pattern, and consistent families exist (ask_pipeworx*, polymarket_*, resolution_*, zyte_*). Minor deviations like noun-only names (polymarket_edges, pipeworx_feedback) and the ask_pipeworx_beta variant are easy to predict, so the pattern is still highly readable.

39 tools is well above the 25-tool threshold for 'too many', and the count is especially unjustified because the server is named Zyte but only 3 of 39 tools are Zyte scraping tools — the rest are a sprawling Pipeworx data, prediction-market, memory, and subscription suite. The surface feels like several servers merged into one rather than a coherent scoped toolkit.

The data-lookup side is broadly covered with universal routing, entity resolution, company profiles, comparisons, claim validation, memory, and subscriptions. However, the Zyte scraping surface (the server's namesake) is thin — only fetch, extract, and screenshot, with no pagination, custom schema, or interaction tools — and odd one-offs like generate_llms_txt and scan_dependency suggest a fragmented domain rather than a complete one.