Skip to Content
What you can doSummarize on-call alerts

Summarize on-call alerts

For on-call engineers woken by a wall of alerts with no context. When an alert fires, the assistant gathers the related metrics and recent logs, compares them with past incidents, and posts a plain-language summary with the most likely runbook to your incident channel. The person on call starts fixing instead of gathering.

What you need

  • Your monitoring and logging tools’ APIs (for example Datadog, PagerDuty or Grafana Loki), and your runbooks somewhere searchable.
  • A Slack app token that may post messages, for the summary.
  • Your alerting tool able to call a web address when an alert fires.

How to build it

  1. Connect your signals. Open APIs and click Connect an API for your monitoring and logging tools, and for your runbook search. For posting the summary, create a tool under Tools, Create tool that calls Slack’s “post message” with your app’s token from Credentials.
  2. Create the assistant. Open Agents, click Create agent and switch to Autonomous. In Instructions, write: “Given an alert, gather the related metrics and the last 15 minutes of logs, compare with earlier incidents, and post a short summary with the likely cause and runbook to #incidents.” Pick the tools under Tools and APIs.
  3. Let it remember. Turn on Remember between conversations with This agent’s own memory, so a recurring incident comes with last time’s fix attached.
  4. Keep it advisory. It summarizes and suggests; give it no tools that change your systems. If you later want it to act, have it ask for approval first (see Approvals).
  5. Trigger it. Point your alerting tool’s webhook at the assistant (see below).

Ideas

For developers

The address to call and sample code are in the Integration card on the assistant’s Overview tab. If your alerting tool cannot send what the endpoint expects, a small relay can sit in between. See the OpenAI-compatible API.

First pass on pull requestsSeveral coding assistants