About this session Model Context Protocol (MCP) has become the standard way for AI assistants to reach real tools, APIs, and data. When MCP servers run on Kubernetes - a common pattern - you need to connect and secure them properly, especially given the level of access and power they might possess. In this hands-on workshop, you'll run MCP servers on Kubernetes and take one from "exposed and wide open" to "connected, observed, and enforced" with the Isovalent toolkit. To make the risk concrete, you'll replay a real MCP server CVE, watch it succeed, then shut it down live, and finish with a hands-on exam that earns you an Isovalent badge. No prior MCP experience is needed. If you're comfortable with kubectl and everyday Kubernetes objects, you're ready. WHAT YOU WILL LEARN Deploy MCP servers on Kubernetes and expose them over stdio and streamable HTTP transports. Explore an MCP server's tools, resources, and prompts live using the MCP Inspector. Connect and control access to MCP servers with Cilium network policy. Observe MCP traffic and in-container process activity with Hubble and Tetragon. Replay a real MCP server CVE and shut it down with a Tetragon runtime policy. Prove your skills in a hands-on exam and earn an Isovalent badge on completion. Can't make it to the live session? Register anyway and we'll send you the recording.