Leie

Several tools occupy overlapping functional territory—ask_pipeworx, ask_pipeworx_beta, ask_pipeworx_grounded, and deep_research all answer factual/research questions, and discover_tools overlaps with suggest_questions. The descriptions are detailed and steer selection, but ask_pipeworx_beta is currently identical to ask_pipeworx, and the polymarket edge/arbitrage/research cluster requires careful reading to distinguish.

Names are uniformly lowercase snake_case and mostly follow a verb_noun or domain_noun pattern (e.g. validate_claim, compare_entities, polymarket_arbitrage). Minor deviations exist—entity_profile and company_facts are noun-first, and remember/recall/forget are bare verbs—but the overall convention is predictable and readable.

With 38 tools, the server is over-packed for a single MCP surface. It spans data retrieval, prediction markets, company research, memory, subscriptions, provider exclusion screening, AI visibility, and even one-off utilities like generate_llms_txt and scan_dependency, making it feel like an aggregated platform rather than a focused toolset.

Core workflows are well covered: querying has multiple modes, company research has profile/facts/comparison/change paths, prediction markets have analysis, arbitrage detection, fill-risk, and resolution auditing, and memory/subscriptions have full lifecycles. Minor gaps exist—there is no execution/order-placement for prediction markets and no way to update a subscription—but agents can work around these.