Skip to main content

PagerDuty Incident Indexing

TL;DR
  • Agent picks up PagerDuty incidents within 1–2 minutes of firing
  • Investigates using all connected data sources — Azure Monitor, Kusto, GitHub, and more
  • Acknowledges, resolves, or escalates based on your response plan configuration
  • Resolution metrics tracked automatically — see agent-resolved vs human-resolved trends

The problem: PagerDuty fires, you context-switch

When a PagerDuty alert fires, you're not just reading a notification — you're starting a manual investigation. Open PagerDuty for the incident details, then Azure Monitor for metrics, then your logs for errors, then GitHub for recent deployments. By the time you've gathered enough context to form a theory, 30 minutes have passed.

Meanwhile, MTTR tracking lives in spreadsheets or PagerDuty analytics, disconnected from what the agent learned during past investigations. Knowledge from resolved incidents stays in engineers' heads, lost when the team rotates.

How PagerDuty incident indexing works

When you connect PagerDuty to your agent:

Scanner polls every minute — your agent checks PagerDuty for new incidents matching your response plan filters (priority, service, incident type). Connect with a PagerDuty API access key from your PagerDuty API Access Keys settings.

Priority mapping — PagerDuty priorities (P1–P5) map to agent severity levels (1–5), so your response plans filter consistently regardless of which incident platform you use.

Status sync — PagerDuty incident status (Triggered → Acknowledged → Resolved) maps to agent status. When the agent resolves an incident, it tracks whether it was agent-mitigated, agent-assisted, or human-resolved.

From there, the agent follows the same investigation and response flow as any other incident platform.

What makes this different

Unlike PagerDuty's built-in automation, which runs predefined workflows, your agent reasons about each incident individually. It correlates evidence across all your data sources, forms hypotheses, and validates them — adapting its approach based on what it finds.

Unlike static runbooks, your agent learns from every resolved incident. AI analysis captures root cause summaries and extracted knowledge, building institutional memory that survives team rotation.

Unlike manual investigation, your agent starts within seconds of the alert firing. No context-switching between tools. No waking up at 3 AM to manually correlate logs and metrics.

Before and after

BeforeAfter
AcknowledgmentWait for on-call to see PagerDuty alertAgent acknowledges in seconds
InvestigationOpen 5+ tools, manually correlate dataAgent queries all sources automatically
Time to root cause30–60 minutes of manual work2–5 minutes, automated
Knowledge capturedIn engineer's head, lost on rotationSaved to agent memory
Resolution trackingManual MTTR trackingAutomated: agent-resolved vs human-resolved

PagerDuty priority mapping

PagerDuty priorities map directly to the agent's severity classification:

PagerDuty PriorityAgent SeverityQuickstart Response Plan
P11 (Critical)✅ Autonomous response
P22 (High)
P33 (Moderate)
P44 (Low)
P55 (Informational)

PagerDuty urgency (high/low) is stored for reference but does not affect agent routing. Priority is the primary severity signal your response plans use.

Incident status normalization

PagerDuty StatusAgent StatusWhat happens
TriggeredActiveAgent begins investigation
AcknowledgedActiveInvestigation continues
ResolvedResolvedAI analysis generates root cause summary

When the agent resolves a PagerDuty incident, it records an SREAgent_Resolved tag on the incident in the agent's own database — so you can distinguish agent-resolved from human-resolved incidents in the agent's analytics. The tag is not written back to PagerDuty.

Get started

Tutorial: Set up PagerDuty incident indexing

CapabilityWhat it adds
Incident Response →How your agent investigates and responds to all incident types
Incident Response Plans →Configure autonomy levels for different incident types and priorities
Root Cause Analysis →AI-driven hypothesis formation and evidence validation
Memory →How resolved incident knowledge improves future investigations
Connectors →Data sources your agent uses during investigation
Was this page helpful?