AI agents are becoming a first-class consumer of web content, right alongside browsers and mobile apps. But most CMS architectures were never designed to be queried, reasoned about, or edited by an autonomous agent.

This talk walks through what it actually takes to expose a headless CMS's content graph through an MCP (Model Context Protocol) server: structuring content models so agents can navigate them reliably, handling auth and permissions for agent-initiated writes, versioning and rollback when an agent gets it wrong, and the schema decisions that make the difference between an agent that "sort of works" and one you can trust in production.

Drawing on real integration work in the headless CMS space, attendees will leave with a concrete architecture pattern they can adapt – not a demo, a blueprint.