WordPress.com Lets AI Agents Write and Publish Site Content

Automattic announced Friday that WordPress.com now allows AI agents to create, edit, and publish website content through natural language commands, a move that could reshape how millions of sites are built and maintained across the web.

The update adds 19 new write operations across six content types \u2014 posts, pages, comments, categories, tags, and media \u2014 to WordPress.com\u2019s Model Context Protocol (MCP) server. AI agents such as Claude, ChatGPT, and Cursor can now manage an entire WordPress site through conversation.

What the Update Enables

The expansion of WordPress.com\u2019s MCP server goes well beyond simple content creation. According to Automattic\u2019s official blog post, AI agents connected to the server can now:

• Draft and publish blog posts with formatting, categories, and tags applied through natural language
• Build landing pages and static pages without touching the WordPress editor directly
• Manage and moderate comments at scale \u2014 approving, flagging, or responding in bulk
• Restructure categories and tags across an entire site\u2019s archive
• Fix missing alt text across a site\u2019s entire media library in a single instruction
• Upload and organize media through conversational commands

The practical effect is that a site owner can now open Claude, describe what they want \u2014 \u201cwrite a 500-word post about our spring sale, tag it \u2018promotions,\u2019 schedule it for tomorrow morning\u201d \u2014 and the agent handles every step from draft to publish without the user opening the WordPress dashboard at all.

What Is the Model Context Protocol?

The Model Context Protocol is an open standard, originally developed by Anthropic, that defines how AI models connect to and interact with external tools and data sources. It functions as a universal connector: instead of each AI company building a custom integration for every platform, platforms expose an MCP server and any MCP-compatible agent can interact with it immediately.

WordPress.com launched read-only MCP support earlier this year, allowing AI agents to retrieve content, check post status, and query site data. Friday\u2019s update adds the write layer \u2014 the ability to actually change things \u2014 which is where the practical utility for site owners begins.

WordPress powers an estimated 43% of all websites on the internet. Giving AI agents write access to that ecosystem is one of the largest expansions of agentic capability to reach the consumer web to date.

Why This Shifts the CMS Landscape

Content management systems have historically been optimized for human interfaces \u2014 dashboards, WYSIWYG editors, drag-and-drop builders. The MCP write integration represents a different bet: that the primary interface for maintaining a website, for a growing segment of users, will be a conversation with an AI agent rather than a browser-based GUI.

For individual site owners, the immediate use case is time savings on routine content tasks \u2014 publishing recaps, updating category structures, bulk-correcting accessibility issues. For agencies managing dozens or hundreds of client sites, the potential is more significant: an agent that can execute multi-site content operations from a single instruction set, without logging into each dashboard individually.

The risk that Automattic is absorbing is quality and intent control. Write access means an AI agent with the right credentials and a bad prompt can publish, modify, or delete content at scale. The company did not detail specific guardrails in its announcement, but the MCP standard supports permission scoping \u2014 meaning individual installations can restrict which operations an agent is authorized to perform.

The Competitive Context

Automattic\u2019s move follows a broader race among platform companies to become AI-agent-native. Shopify, Notion, Linear, and GitHub have all shipped or announced MCP server support in recent months, each trying to ensure their platform is a first-class destination for the growing population of users who work primarily through AI agents rather than native apps.

For WordPress specifically, the timing carries additional weight. The platform has faced mounting competition from newer website builders \u2014 Webflow, Framer, Squarespace \u2014 that have marketed themselves as more modern alternatives. Positioning WordPress.com as the most AI-capable CMS on the web, at a moment when agentic workflows are becoming a mainstream expectation for technical users, is a credible strategic counter.