MCP on WordPress
How to get started with MCP on WordPress.com
Getting started only takes a few minutes. Once enabled on your WordPress.com account, MCP works behind the scenes to connect your WordPress.com sites with your favorite AI assistant:
- Enable MCP on your WordPress.com account.
- Configure your AI application following the provided instructions.
- Authenticate through our OAuth interface.
After that, you can directly ask your AI assistant for information about your sites so you donโt have to dig through reports for basic answers.
Here are just a few examples of some of the things you may want to learn about your sites through your AI assistant:
- โHow did my latest post perform?โ No need to click through pages of analytics; you get the traffic insights instantly in your AI assistant.
- โRun a comprehensive health check across all my WordPress.com sites โ show me which ones need attention for storage, performance, or content freshness, then create a prioritized action plan.โ Your AI assistant can instantly return a prioritized action plan โ no manual site-by-site checks required.
- โSummarize recent comments across all my sites.โ Address common reader questions and needs by getting a quick overview of the conversations happening across your sites.
These are just a handful of ways MCP makes your WordPress.com site AI-readable. See the complete list of available MCP tools and some prompt examples in our developer documentation.
Enable MCP tool access for AI assistants
https://wordpress.com/support/model-context-protocol-mcp-settings/?ref=blog
The support page for WordPress.com’s Model Context Protocol (MCP) settings details how users can connect their WordPress sites to AI assistants (like Claude, Cursor, or VS Code) to provide them with real-time site context.
Here is a summary of the key points from the guide:
1. What is MCP on WordPress?
MCP is an open standard that allows AI models to securely “read” your WordPress data. Instead of copying and pasting information into a chat, the AI can query your site directly to understand your content, comments, and settings before generating a response.
2. Key Capabilities (Current Status)
- Read-Only Access: Currently, the integration is strictly read-only. AI assistants can view your posts and settings to provide information, but they cannot create, edit, or delete content.
- Privacy & Consent: Data is only shared when you give the AI explicit permission to use a tool. WordPress.com does not use this data to train AI models; it is used solely to provide context for your specific request.
3. How to Enable MCP
To activate the feature, users must:
- Navigate to Account Settings > MCP (or
wordpress.com/me/mcp). - Toggle “Enable MCP Access” to the ON position.
- Click “Configure MCP client” to get the necessary connection details for your AI application (e.g., Claude Desktop).
4. Management and Granular Control
WordPress provides high-level control over what the AI can see:
- Disable Specific Tools: You can toggle individual capabilities off (e.g., you might allow an AI to see “Posts” but disable access to “Comments”).
- Site-Specific Settings: If you manage multiple websites, you can enable or disable MCP access on a per-site basis.
5. Why Use It?
The primary benefit is turning your CMS into a Context Management System. This allows you to ask an AI questions like:
- “Summarize the feedback from my recent comments.”
- “Which of my posts are missing SEO descriptions?”
- “Based on my existing writing style, suggest three new blog topic ideas.”
WordPress MCP Extension
[Gemini]
In the context of modern web development, a Content Management System (CMS) like WordPress is evolving into a Context Management System. This shift means that instead of just serving pages to humans, WordPress can now serve structured “context” to AI models to help them understand and manage your site.
The Model Context Protocol (MCP) is the open standard that makes this possible.
1. From Content to Context
Traditionally, WordPress manages content (text, images, metadata). However, for an AI (like Claude or ChatGPT) to be useful, it needs contextโit needs to know your siteโs categories, recent posts, user permissions, and active plugins.
- Content Management: Storing a blog post so a visitor can read it.
- Context Management: Exposing that postโs structure, SEO data, and site settings so an AI can “reason” about it and perform tasks.
2. What is the MCP Extension?
The Model Context Protocol (MCP), introduced by Anthropic, acts like a “USB port” for AI. It allows an AI agent to “plug into” WordPress.
In WordPress, this is usually implemented via a plugin (like Automatticโs MCP Adapter or the built-in server on WordPress.com) that translates between the AIโs language and the WordPress REST API.
How the Architecture Works
| Component | Role in WordPress |
|---|---|
| MCP Client | The AI interface you use (e.g., Claude Desktop, Cursor, VS Code). |
| MCP Server | The WordPress plugin or service that “speaks” MCP. |
| Tools | Actionable commands the AI can run (e.g., create_post, list_users, update_settings). |
| Resources | Data the AI can read (e.g., specific post content, media library files). |
3. Practical Use Cases
By extending WordPress with MCP, you transform the CMS into an AI-powered workstation:
- Conversational Admin: Instead of clicking through menus, you can tell your AI, “List all pending comments and mark the ones that look like spam as trashed.”
- Automated Content Creation: “Draft a 500-word post about our new product using the tone of our last three announcements.” (The AI reads previous posts via MCP to match the style).
- Site Audits: “Check the meta descriptions of my top 10 posts and suggest improvements for SEO.”
- WooCommerce Management: “Show me all orders from the last 24 hours that haven’t shipped and draft an update email for the customers.”
4. Security and Control
A common concern is giving AI “the keys” to your site. MCP includes strict guardrails:
- Read-Only vs. Read/Write: You can configure the server to only let the AI see data without being able to change it.
- Permissions: The AI typically operates under a specific user account; if that user can’t delete a post, the AI can’t either.
- Human-in-the-loop: Most MCP clients require you to approve a “tool use” (like publishing a post) before it actually happens.
Next Steps
Would you like me to walk you through how to install and configure an MCP server on your own WordPress site? The video below provides a practical walkthrough on how to set up the WordPress MCP plugin and connect it to AI tools like Claude to automate development and management workflows.
