Technographics

Several tool clusters overlap: ask_pipeworx, ask_pipeworx_beta, and ask_pipeworx_grounded all route the same 6,402 tools (beta is currently identical to stable), and five polymarket_* tools all surface 'opportunities' via different mechanisms. The verbose descriptions do a good job delineating when to use each, but the boundaries require careful reading and misselection is easy.

The dominant pattern is snake_case verb_noun (validate_claim, compare_entities, tech_detect, list_subscriptions) with consistent brand/domain prefixes for families (polymarket_*, pipeworx_*). A handful of bare verbs (remember, forget, recall, subscribe, unsubscribe) and the brand-style ask_pipeworx (no underscore) deviate slightly, but the overall scheme is predictable.

At 38 tools, this exceeds the 25+ 'too many' threshold, and the surface spans very different domains (data lookup, prediction markets, AI visibility, web tech detection, memory, subscriptions). It feels like multiple servers' worth of functionality bundled together, making the option set heavy for an agent to navigate.

Core workstreams are deeply covered: memory and subscriptions have full CRUD lifecycles, prediction markets are extremely thorough (research, arbitrage, edge tracking, fill risk, cross-venue spread, resolution audit), and company research spans profile, facts, comparison, changes, and verification. Minor gaps exist (no explicit pipeworx:// record fetcher; a few isolates like generate_llms_txt) but nothing causes dead ends.