Incident Response Agent · Use case
How can AI triage PagerDuty incidents with Grafana and Prometheus data on our own infrastructure?
An incident response agent reads an alert in PagerDuty, pulls the related metrics and logs from Grafana, Prometheus, Dynatrace or CloudWatch, and drafts an incident summary with a likely cause and the runbook that applies. It proposes and an engineer acts: the agent does not carry out a runbook step. It runs on your own infrastructure, so operational data stays inside.
An alert fires.
The gathering starts by hand.
- ✗ 01When an alert fires, the on-call engineer moves between monitoring screens to gather metrics, logs and past incidents before deciding anything.
- ✗ 02Operational data describes the inside of the organisation's systems, so it cannot be sent to a hosted assistant.
Where it stops for a person
Before any runbook step.Before any runbook step. The agent names the step it would take and why, and an engineer carries it out.
It gathers and proposes.
An engineer acts.
Step · 01
The agent reads the alert in PagerDuty.
Step · 02
It pulls the metrics and logs for the affected service from the monitoring tools.
Step · 03
It looks for similar past incidents in PagerDuty and ServiceNow.
Step · 04
It drafts a summary: what is failing, the likely cause and the runbook that applies.
Step · 05
The on-call engineer reads the summary and decides what to do.
Reads from · connectors
The agent connects to PagerDuty and ServiceNow through connectors, for alerts and past incidents.
Acts through · MCP servers
Grafana
Prometheus
Dynatrace
CloudWatch
Monitoring servers. The agent connects to Grafana, Prometheus, Dynatrace or CloudWatch through MCP servers to query metrics and logs.
What you set up
- Connect PagerDuty and your monitoring tools, with read access.
- Give the agent your runbooks.
- Write test incidents with known causes to score it before release.
Where it runs
The agent, its knowledge store and the model run on infrastructure you own or control. You choose the model and serve it yourself, so questions, documents and answers are not sent to a public AI service.
How you get it today
This page is labelled Use case: the agent is built in Katonic Studio. You describe the job, connect what is named above, add test questions and release the agent once it passes them. When the Catalog holds a tested template for this job, this page will carry the Template label and show how to take it.
Can the agent restart a service or roll back a release?
No. It gathers, summarises and proposes. An engineer carries out any runbook step.
Which monitoring tools does it connect to?
Grafana, Prometheus, Dynatrace and CloudWatch, through MCP servers. Alerts and past incidents come from PagerDuty and ServiceNow.
Does it update the incident in PagerDuty or ServiceNow?
No. It reads from both. The engineer records the outcome.
Where does operational data go?
Nowhere outside. The agent and the model run on your own infrastructure.