Related Servers

The descriptions aggressively cross-reference each other (check_target vs check_now, pause_monitor vs create_maintenance_window, list_monitors vs list_heartbeats vs list_check_results), so nearly every pair has an explicit boundary. The residual risk is in clusters of near-siblings: get_incident / get_incident_context / get_incident_diagnosis, and the four status-page-incident writers (draft / create / post_update / update), where an agent must read the description carefully to pick correctly. Nothing is truly duplicated, but several tools are one misread away from the wrong call.

Almost everything is a clean snake_case verb_noun (list_monitors, create_heartbeat, delete_maintenance_window, bulk_update_monitor_tags), with bulk_ and bulk_*_tags variants following the same pattern. Minor deviations: list_active_maintenance drops the noun used by its sibling list_maintenance_windows, and post_status_page_update / verify_deploy / check_now use less parallel verbs than their neighbours. Still highly predictable overall.

56 tools for an uptime-monitoring API is heavy: five bulk_* variants, three incident-read tools, three uptime/response-time readers, and helper endpoints like get_exposed_ports_preset and list_active_maintenance that are arguably data on other tools. Each tool is documented and most are defensible, but the set crosses well past the point where an agent can hold the surface in view, and some tool calls are only needed for filtering conveniences.

Coverage is unusually thorough: full lifecycle for monitors, heartbeats, maintenance windows, tags and status-page incidents, plus read/diagnostic layers (incidents, diagnosis, check results, uptime summary/daily, per-region latency, changes, deploy record/verify, plan usage, regions, channels). The known gaps — status pages and channels are dashboard-only, no bulk delete, incidents resolve automatically — are explicitly documented as intentional rather than silent dead ends.