Large language models are getting good at using tools — but they need a bridge to your CMS. The Model Context Protocol (MCP) is that bridge, and plone-mcp is an open-source MCP server that lets AI clients like Claude Desktop manage a Plone site through its REST API. We'll cover what becomes possible when an LLM can create and edit content, assemble Volto blocks, run searches, drive workflow transitions and do translations — all through natural language. After a live demo (building multilingual pages, composing block layouts, publishing), we'll open up the architecture: how the server maps MCP's tools/resources/prompts primitives onto the Plone REST API, key design decisions like the block-layout preparation pattern, and where the rough edges still are. You'll leave knowing what MCP is, how it fits Plone, how AI-assisted editing could reshape content work — and how to try it and contribute.