Check diagnostics

All four operations require checks:history and an owned check ID. Paths are relative to /api/v1. days accepts seven or 30, default 30. MCP uses the same parameters in a tool argument object.

Daily history

GET /checks/:id/history / get_check_history.

Returns id, start_at, end_at, retention_days, and days, an array of:

Daily field	Meaning
date	UTC calendar date
run_count	Recorded activity count
average_duration_ms	Mean measured duration or null
incident_count	Incidents starting on that date

The window starts at UTC midnight days-1 calendar days before today, ending now. An empty day is still represented; absence of measurements is not zero latency.

Retained runs

GET /checks/:id/runs / list_check_runs. Accepts days, limit, cursor.

Returns id, start_at, end_at, items, next_cursor. Each run has: id, received_at, nullable duration_ms, nullable probe_id, and nullable resolver_id. Runs are paged by underlying ID, not by descending time.

No payloads, sender IPs or credentials are returned. Retention limits samples even if the requested window is longer.

Latency statistics

GET /checks/:id/metrics / get_check_metrics. Website, DNS and TCP only; other kinds return invalid_kind. Accepts days and at most one of probe_id or resolver_id. Use IDs from this check’s run results; foreign sources are not accepted.

Result	Meaning
id, start_at, end_at	Check and UTC calendar window
run_count	All selected recorded runs
measurement_count	Runs with nonnegative measured duration
probe_id, resolver_id	Selected source or null
min_duration_ms, average_duration_ms	Minimum and arithmetic mean, or null
p50_duration_ms, p95_duration_ms	Percentiles of available durations, or null
max_duration_ms	Maximum measured duration or null

Statistics pool selected samples, not per-location averages. They are not availability estimates. A run can exist without a latency measurement.

Uptime

GET /checks/:id/uptime / get_check_uptime. Accepts days.

Returns id, start_at, end_at, window_seconds, downtime_seconds, nullable uptime_percentage, basis: "incident_free_time" and paused_and_pending_history_available: false.

This is a rolling days-long interval ending now, clipped to check creation. Downtime merges/clips overlapping incidents, including ongoing episodes. Degraded and down both contribute incident time.

No recorded activity or incident evidence means the API percentage is null. Historical paused/pending periods cannot be reconstructed; do not label this as continuously verified uptime through those periods.

curl --fail-with-body --silent --show-error \
  --header "Authorization: Bearer $PINGSTACK_TOKEN" \
  "$PINGSTACK_BASE_URL/api/v1/checks/$PINGSTACK_CHECK_ID/uptime?days=7"

Set PINGSTACK_CHECK_ID from a check response. A daily count, latency distribution and incident-free percentage answer different questions.