Investigate Alert

Investigate a monitoring alert by pulling metrics, logs, traces, and related service code. Symptoms in, root cause hypothesis out.

When to Use

• A monitoring alert fired and you need to understand why
• On-call engineer needs a structured investigation starting point
• Alert is noisy and you want to determine if it's actionable or a false positive

Process

1. Verify Monitoring MCP Availability

Check which monitoring MCP servers are available. Look for any mcp__* tools related to monitoring platforms (Datadog, Grafana, PagerDuty, etc.).

Recommended: Datadog MCP — provides the richest investigation surface (monitors, metrics, logs, traces, events in one platform).

If no monitoring MCP is available, stop with: