Noaa Tides

Several large clusters are hard to tell apart: ask_pipeworx, ask_pipeworx_beta, ask_pipeworx_grounded, and deep_research all answer factual questions with nearly identical routing (the beta is explicitly stated to currently match the stable router), and polymarket_arbitrage, polymarket_edges, polymarket_edge_tracker, polymarket_fill_risk, and polymarket_kalshi_spread form an overlapping prediction-market family. The NOAA tools themselves (water_level, currents, met_obs, predictions, etc.) are distinct, but the meta/query families create real misselection risk.

Nearly all names use consistent snake_case with clear verb_noun or domain_noun patterns (list_subscriptions, generate_llms_txt, polymarket_fill_risk, scan_dependency). The main deviation is a handful of bare single nouns (currents, datums, predictions, stations, recall, remember) mixed alongside the verb-led names, which is readable but not perfectly uniform.

38 tools is very heavy, and only ~7 of them belong to the stated NOAA Tides domain; the rest are a broad general-purpose Pipeworx gateway surface (research, prediction markets, entity resolution, memory, feedback). This is a severe mismatch between the server's apparent scope and its actual tool count.

For the NOAA Tides & Currents domain the coverage is strong: station listing, station metadata, water levels, currents, predictions, datums, and meteorological observations together cover the core CO-OPS surface with no major dead ends. Minor gaps (e.g. harmonic constituents or derived products) keep it from a full 5.