Monitoring Agent Pings the On-Call Dev's Claude Code When a Service Degrades

It's 3am. Error rates spike on the checkout service. PagerDuty fires. The on-call dev wakes up, opens their laptop, stares at dashboards, tries to figure out what changed. Twenty minutes of orientation before they can even start fixing anything.

Pre-investigated incidents

The monitoring agent detects the spike and messages the on-call developer's Claude Code with the details: which service, which endpoints, error rates, when it started, and what changed recently (recent deploys, config changes, dependency updates). The developer's Claude Code starts correlating: it checks the git log for recent changes to the affected service, looks at the error patterns, and builds a preliminary incident report.

When the developer opens their laptop, they don't start with "what's happening?" They start with "your checkout service error rate jumped 40% at 2:47am, likely caused by the database migration in commit f82a3d that added a NOT NULL column without a default value. Here's the relevant code and a suggested rollback."

Time to resolution drops dramatically when the investigation starts immediately, not when a human finishes rubbing their eyes.