Agentic Incident Response
summary
An AI agent automatically triages alerts, queries Grafana, Loki and Prometheus, checks recent deployments and the code repo, and produces a data-backed root cause with recommended actions; example finds include a SQL typo causing "no such column due" and a memory leak where the agent patches Kubernetes to raise a pod memory limit to 500MB and restores service. Implementation uses the model context protocol with tools for logs/metrics and is read-only by design until permissions escalate.
AI-written summary · about
chapters (11)
- 0:00Agentic incident response intro
- 1:06AI-powered incident vision
- 8:38Why agentic response now
- 10:05AI agents vs LLMs
- 12:51Agents using tools and prompts
- 13:58Model Context Protocol role
- 16:33AI responder under the hood
- 19:02Live AI incident investigation
- 24:44Fixing memory leak demo
- 29:20Summary and AI future
- 31:00Q&A on metrics and access
related talks
30mAI Agents for Incident Management
34mAgentic Incident Response
15mIntelligent Incident Management in DevOps
40mWhen Incidents Fix Themselves: AI SRE in action
19mFrom Alerts to Action: AI-Powered Incident Response Systems
16mAI Agents for Incident Investigation: The Good, The Bad, and The Ugly
19mAI Agents for Incidents: The Good, the Bad, and the Ugly
20mAutomated Incident Management