Incident Triage Agent — DevOps skill for Claude Code

Autonomous Incident Triage Agent

An autonomous AI agent that monitors system alerts, correlates related signals, identifies root causes, and automatically creates structured incident tickets.

**Built with:** AWS Bedrock, Claude via Bedrock, Strands Agents SDK

Setup

python -m venv venv
source venv/bin/activate
pip install strands-agents boto3

Run

# Local logic test (no API)
python test_logic.py

# Full agent with Bedrock (requires AWS credentials)
python main.py

How it works

1. Fetches recent alerts from monitoring
2. Groups alerts by timing & service dependencies
3. Pulls metrics to confirm hypotheses
4. Creates incident tickets with evidence-based root cause analysis

Files

• mock_alerts.py — Sample alert & metric data
• tools.py — Agent tool definitions
• main.py — Strands agent with Bedrock integration
• test_logic.py — Local logic verification (no API)

Demo Output

The agent correctly:

• Groups A1, A2, A3 as one incident (checkout-api + checkout-db DB pool exhaustion)
• Isolates A4 as noise (unrelated auth-service event)
• Cites metric evidence: db_pool_used: 98%, connections: 98/100
• Assigns high severity due to connection pool type