incidents

Incidents

Preflight — read first. If you cannot see any uptimerobot:* MCP tools in your tool list, invoke the uptimerobot:setup skill before doing anything else. Do not tell the user the MCP is misconfigured — setup's Step 0 detects the common case (server connected, tools loaded after session start) and resolves it without re-keying.

Two tools:

• list-incidents — paginated incidents across the account or for one monitor.
• get-incident-details — checker locations, IPs, logs, traceroute for one incident.

List incidents

{
  "timeRange": "7d",
  "limit": 25
}

Parameters:

• timeRange — "24h" / "7d" / "30d" (1h–90d), or ISO 8601 interval "2024-01-01T00:00:00Z/2024-01-31T23:59:59Z". Defaults to "7d".
• monitorId — optional, restrict to one monitor.
• cursor, limit — pagination (follow the instructions field for the next call).

Filter by monitor

{
  "monitorId": 800123456,
  "timeRange": "30d"
}

Absolute time range

{
  "timeRange": "2025-04-01T00:00:00Z/2025-04-15T23:59:59Z"
}

Get incident details

{ "incidentId": "inc_abcdef1234567890" }

Note: incidentId is a string, not a number.

Response includes:

• Monitor id and friendly name.
• Start / end timestamps, duration.
• Root cause category (if assigned).
• Per-checker probe results: checker location, IP, HTTP status code observed, error kind, response body snippet.
• Traceroute hops where available.

Use this to answer "why did this go down?" — show the user the checker locations first, then the error types, then (if asked) the verbose logs.

Common mistakes

• Passing a numeric incidentId. It's always a string.
• Requesting ranges longer than 90 days — the server rejects the call. For older data, the user needs to query it in the UptimeRobot dashboard.
• Forgetting that list-incidents is paginated. Loop with the cursor from instructions until hasMore: false.
• Inventing incidentIds. Always source them from list-incidents.

Output tips

• Group incidents by monitor before summarising.
• Lead with active (unresolved) incidents, then recent resolved, then older.
• Surface total downtime duration, not just incident counts — the user usually cares about impact.
• Don't paste raw log bodies unless the user asks. Summarise the error kind and HTTP status.

Related

• stats — for aggregated downtime minutes across a time range.
• manage-monitors — to pause a flapping monitor after confirming incidents.