Skip to main content

Incident platforms

An incident platform is the system that tells your agent when something goes wrong. By connecting your incident platform, your agent can receive alerts, investigate issues, and take action—automatically, without waiting for someone to start a chat.

Flow showing incident platform sending alerts through response plans to agent investigation and action

Without an incident platform, your agent is reactive: users ask questions and it investigates on demand. With one connected, your agent becomes proactive: it picks up incidents the moment they fire and starts working immediately.

Supported platforms

PlatformWhat it provides
Azure MonitorConnected by default — alerts from your managed resource groups flow to the agent automatically
PagerDutyIncident alerting and on-call management with API-based integration
ServiceNowEnterprise IT service management integration

Only one incident platform can be active at a time. Azure Monitor is the default — switching to another platform disconnects Azure Monitor.

What connecting an incident platform enables

Once connected, your agent gains these capabilities:

Automatic incident reception

Incidents flow to your agent the moment they're created in your platform. No one needs to copy-paste alerts or manually start an investigation—the agent picks up incidents automatically.

Rich incident cards

Incoming incidents from all supported platforms — PagerDuty, ServiceNow, and Azure Monitor — display as rich cards in the chat interface. Each card shows:

FieldDetails
Severity badgeColor-coded by priority (e.g., P1/Sev0 = red, P2/Sev1 = orange)
TimestampWhen the incident fired
TitleIncident title with platform prefix
StatusCurrent status (e.g., Triggered, Acknowledged)
DescriptionIncident summary
Response planLink to the response plan handling the incident (if configured)
View DetailsLink to the incident in its source platform

Rich cards replace the plain-text incident notifications used previously, making it easier to scan incident details at a glance.

Incident interaction

Your agent can read and write back to the incident. These tools are available automatically when the corresponding platform is connected — no additional setup needed.

PlatformRead capabilitiesWrite capabilities
Azure MonitorAlert details, severity, affected resourcesAcknowledge alerts, close alerts
PagerDutyIncident details, diagnosticsAcknowledge, resolve, add notes
ServiceNowIncident detailsPost discussion entries, acknowledge, resolve

Response plans

Response plans define what your agent does when specific types of incidents arrive. You configure rules based on incident severity, title patterns, or other criteria, and the agent follows the plan automatically.

Learn more: Incident Response Plans →

Diagram showing how response plans combine filters, autonomy levels, and custom instructions

A response plan can:

  • Run specific investigation steps
  • Use particular connectors and tools
  • Operate at a defined autonomy level (from "gather info only" to "take corrective action")
  • Retry investigation automatically (up to a configurable limit) before escalating to a human

Response plans turn your agent from a general-purpose assistant into an incident responder with defined procedures for known incident types.

Quickstart response plan

When you connect an incident platform, you can enable Quickstart response plan to automatically create a default response plan. This gets you started immediately:

PlatformDefault plan handlesAutonomy level
Azure MonitorSev0, Sev1, Sev2 alertsAutonomous
PagerDutyP1 incidentsAutonomous

Azure Monitor supports all severity levels (Sev0–Sev4). The quickstart plan targets the highest-priority alerts by default — you can customize it to include additional severities or create separate plans for lower-priority alerts.

The quickstart plan creates a response plan named quickstart_handler that:

  • Matches incidents by priority/severity
  • Covers all impacted services
  • Runs in fully autonomous mode
  • Can be customized or disabled later

You can customize this default or create additional response plans with different filters and autonomy levels.

Track incident value

The Monitor → Incident metrics section gives you visibility into how your agent handles incidents over time.

Learn more: Track Incident Value →

KPI cards showing incidents reviewed, mitigated by agent, assisted, and pending

MetricWhat it shows
Incidents reviewedTotal incidents processed by the agent
Mitigated by agentIncidents the agent resolved autonomously
Assisted by agentIncidents where agent helped, user completed resolution
Mitigated by userIncidents resolved by user with agent-provided information
Pending user actionIncidents waiting for human input

Use these metrics to understand your agent's effectiveness and identify response plans that might need tuning.

Incident platforms vs. connectors

These are different concepts that work together:

Incident platformsConnectors
PurposeWhere alerts come FROMData and actions agent can USE
Configured inBuilder → Incident PlatformBuilder → Connectors
DirectionInbound (incidents flow TO agent)Outbound (agent reaches OUT to systems)
ExamplePagerDuty sends an alert → agent investigatesAgent queries Kusto → finds root cause

Your agent uses both: the incident platform triggers the investigation, and connectors provide the tools to investigate.

ResourceWhy it matters
Tutorial: Set up response plans →Step-by-step guide to create your first response plan
Incident Response Plans →How response plans route incidents to custom agents
Automate Incident Response →End-to-end incident automation capabilities
Track Incident Value →Measure your agent's incident resolution impact
Monitor Agent Usage →Track usage, session insights, and agent activity
PagerDuty →PagerDuty-specific setup and capabilities
ServiceNow →ServiceNow-specific setup and capabilities
Connectors →How connectors provide tools for investigation
Was this page helpful?