WordPress.com now allows AI agents to draft, edit, and publish posts on customer sites through MCP-powered natural language commands.
WordPress.com announced AI agents can now draft, edit, publish content, manage comments, update metadata, and organize content with tags and categories on customer websites. This builds on MCP (Model Context Protocol) support introduced last fall, which previously allowed AI assistants only read access. The interface lets site owners issue natural language commands to control agents. WordPress.com reaches 20 billion page views and 409 million unique visitors monthly, amplifying the web-scale implications of this move.
WordPress.com's MCP integration is a live, production-grade example of agentic write access at scale — not a demo. Agents can now perform CRUD operations on one of the internet's largest publishing platforms through natural language, using MCP as the protocol layer. This sets a concrete architectural precedent: MCP as the standard bridge between LLMs and CMS platforms, which means the tooling you build today on MCP will have direct real-world analogues.
Connect Claude Desktop to a WordPress.com site via MCP this week and test agentic post drafting and publishing — benchmark the round-trip latency and error rate to evaluate whether MCP is production-ready for your own platform integrations.
Open Claude Desktop, connect it to a WordPress.com account via its MCP server, and issue this command: 'Draft a 300-word post titled Test Post, add the tag AI-generated, and save it as a draft.' Check WordPress.com's post editor to confirm the draft appeared with correct metadata.
Tags
Signals by role
Also today
Tools mentioned