From 4 Hours to 8 Minutes with AI Agents That Transform SRE...
summary
Claims a workflow that cuts a typical SRE investigation from four hours to eight minutes by chaining specialized agents. It builds three concrete agents — Terraform agent, GitOps agent, infrastructure validation agent — wired through MCP and KAgent on Kubernetes, using Prometheus, Grafana and GitHub tools and models like GPT‑4.1 mini and Gemini 2.5 Pro. The live demo deploys MCP servers, auto‑queries metrics, creates a GitHub issue, generates a Grafana dashboard and successfully edits a broken service YAML to fix the failure.
AI-written summary · about
chapters (12)
- 0:00AI agents for SRE
- 1:01Incident response with AI
- 3:29Reliability engineering framework
- 5:57How AI agents work
- 8:51Agent loop and tool calling
- 15:34Model Context Protocol
- 19:29Multi-agent scenarios
- 20:33Building AI agents
- 23:26Live demo setup
- 31:21SRE agent health analysis
- 37:42AI fixing service issues
- 41:40Q&A
related talks
40mWhen Incidents Fix Themselves: AI SRE in action
36mSRE'ing the SRE AI Agents
34mAutomating SRE Operations with Multi-Agent AI
19mMentoring the Machine
30mArchitecting Integrity: Responsible Software Evolution in the AI Era
30mAI Agents for Incident Management
33mStrategic AI Agents for K8s SRE: Beyond Ad-hoc Prompting
40mHow To Tame Your Agents?