MCP Solved Integration. It Didn't Solve Governance. aka What Happens When Cursor Writes to Prod
MCP gave AI agents a universal way to connect to every tool your company runs — databases, cloud infrastructure, internal APIs. The integration problem is solved. But nobody asked who's watching what these agents actually do once they're connected. We've spent the last year building governance infrastructure for MCP at enterprise scale, and what we've seen in the wild is sobering: agents with broad access, no audit trails, and zero policy enforcement between "the LLM decided" and "it happened in production." In this talk, I'll share real examples of what breaks when MCP runs ungoverned, why enterprises keep stalling adoption without a control plane, and what we've learned shipping the tooling to fix it. If you're building with MCP or letting your team use AI coding agents, this is the talk where you find out what you're not seeing.
