Related Servers

Several clusters of tools have nearly indistinguishable boundaries: three earthquake feeds (turkey_earthquakes_afad, europe_mediterranean_earthquakes_emsc, earthquake_monitor_usgs_emsc_afad), six Greenhouse hiring feeds split only by vertical (AI/ML, fintech, cybersecurity, e-commerce, B2B tech), and at least five IP/port/CVE scanners (cloud_perimeter, shodan_internetdb, vpn_gateway_firewall_cve, cyber_cve_kev_epss, ot_ics_scada, b2b_cyber_insurance). Identical boilerplate descriptions ('Real-time structured data extraction feed for X') give the agent no disambiguating signal beyond the long tool name itself.

All names are snake_case and descriptive, which is superficially consistent, but the ordering convention is mixed: some lead with the source brand (eksi_sozluk_scraper, hepsiemlak_scraper, shodan_internetdb_...), others lead with the domain or region (turkey_fuel_prices, global_product_recalls, europe_mediterranean_earthquakes). No verb is used anywhere, so the name conveys the data source rather than the action, leaving agents to infer that everything is a read/extract.

59 tools is far above the practical ceiling for a single agent-facing tool surface; selection becomes a scanning problem rather than a reasoning one. Each tool nominally maps to a distinct upstream data feed, so it is not pure duplication, but the count is well past the point where a well-scoped set would be separable into coherent sub-collections.

The domain is 'read-only real-time intelligence feeds', and for that domain the surface is broad and self-consistent: every feed is an extraction endpoint, so create/update/delete gaps are not meaningful. The one real gap is the absence of any discovery, aggregation, or cross-feed correlation tool to help an agent navigate 59 sibling endpoints.