The Model Context Protocol (MCP) is quickly becoming the backbone of agentic AI, the standard for connecting agents to tools, APIs, and enterprise systems. It’s already inside VS Code, Claude Desktop, and countless custom agents. But here’s the catch: Your AI might be making connections you don’t know about. As MCP adoption explodes, developers are spinning up servers ad-hoc, with inconsistent policies, weak credential handling, and zero visibility. Every new “friend” your agent makes is a potential attack surface. In this session, we’ll unpack the power and risk of MCP in enterprise environments and show how to scale agentic development safely: centralizing security, enforcing access controls, and keeping governance consistent across every connection. Because the last thing you need is your AI making friends behind your back.