Posthog

The set contains several overlapping tool families: ask_pipeworx, ask_pipeworx_beta (explicitly identical today), ask_pipeworx_grounded, deep_research, discover_tools, and suggest_questions all serve lookup/research purposes. There is also heavy overlap among entity/company tools (entity_profile, company_facts, recent_changes, compare_entities, validate_claim) and among prediction-market tools (bet_research, polymarket_edges, polymarket_arbitrage, polymarket_kalshi_spread, polymarket_edge_tracker). Detailed descriptions help, but the boundaries are not crisp enough to avoid misselection.

Most names are snake_case and reasonably readable, with useful family prefixes like polymarket_*, pipeworx_*, and get_*. However, verb conventions are inconsistent: bare verbs (remember, recall, forget), noun-phrase names (ai_visibility_check, release_calendar_markets), and the unusual generate_llms_txt break the pattern.

39 tools is far too many for a server named Posthog, especially since only three tools (get_events, get_insights, get_persons) actually address PostHog functionality. The rest form a sprawling Pipeworx/prediction-market/memory/subscription toolkit that would be more coherent split into separate servers.

For a PostHog product-analytics server, the surface is severely incomplete: there are three read-only getters and no create, update, delete, query, dashboard, cohort, or feature-flag operations. Even as a general data toolkit, the tools do not form a coherent lifecycle around any single domain.