Sunrisesunset

The descriptions work hard to differentiate tools ("START HERE", "PREFER OVER"), but there is a dense cluster of overlapping research entry points: ask_pipeworx, ask_pipeworx_beta, ask_pipeworx_grounded, deep_research, entity_profile, compare_entities, recent_changes, bet_research, and validate_claim all route to the same underlying catalog. ask_pipeworx_beta is explicitly stated to be identical to ask_pipeworx right now, a genuine duplicate an agent cannot distinguish on merit. get_times vs get_times_date also overlap for the 'today' case. The polymarket_* family, memory trio, and subscription trio are otherwise well-separated.

Everything is snake_case, which is good, but the structural conventions mix: verb_noun (compare_entities, resolve_entity, validate_claim), verb_object (get_times, list_subscriptions), bare imperatives (remember, forget, recall, subscribe, unsubscribe), and noun/prefixed forms (entity_profile, polymarket_edges, ask_pipeworx). The polymarket_* and ask_pipeworx families are internally consistent, but there is no single predictable pattern across the set.

33 tools is heavy even for a broad data gateway, and the surface spans unrelated domains (solar times, AI-marketing audits, npm dependency checks, prediction-market arbitrage, memory, subscriptions). That breadth suggests kitchen-sink accumulation rather than a well-scoped set, and the server name 'sunrisesunset' mismatches most of what it actually exposes.

Coverage is genuinely broad: memory has full lifecycle (remember/recall/forget), subscriptions have create/list/cancel/alerts, and research, entity resolution, comparison, and verification are all present with citation URIs. Minor gaps exist (no direct quote/price tool outside the router, edge tracking depends on prior snapshots), but an agent can work around them via ask_pipeworx.