Generative AI is often limited to the IDE. But what happens when the pager rings at 3 AM? In this session, we build an autonomous SRE Agent that reacts to live site incidents. Triggered by alerts, this agent uses the Model Context Protocol (MCP) to investigate Kubernetes clusters, analyze logs, and even execute remediation scripts. We will explore the architecture of "Agentic Operations"—deciding when to let the AI fix the production issue and when to just let it draft the post-mortem.