MCP servers that can list incidents

What to know

All four return the live incident set, but they sit at different points in the stack. PagerDuty is the alerting and on-call layer, so list_incidents there reflects what paged a human and who it routed to. Grafana's list_incidents reads from Grafana Incident, tied to the dashboards and alert rules already in the same workspace. Better Stack pairs incidents with uptime monitoring, and uptime_list_incidents accepts filtering options, so the agent can narrow to a status or window instead of paging the whole history. Datadog's search_datadog_incidents takes a query string, so the agent can scope to active incidents or a severity and read them next to the metrics and logs that triggered them. A read like this is safe to hand an agent freely: it changes nothing, and the worst case is a stale snapshot.

Listing is a snapshot, and an agent that only sees the snapshot keeps rediscovering the same ground. One that remembers an incident on this service last week, or that a given alert is a known false positive the team agreed to mute, writes a triage summary that distinguishes a recurrence from a first-time page. The incident API hands back what's open; the judgment about whether it's noise comes from what happened before.