MCP is how AI assistants like Claude and ChatGPT stop chatting and start working on your WordPress site. Short, honest answers to what everyone actually asks.

New here? MCP for WordPress is the gentle intro. Below, the real questions, answered straight.

The Basics

What even is MCP?

USB-C for AI. It is an open standard that lets an assistant actually do things on your site instead of just talking about them. One port, every tool.

What can it actually do on my site?

Anything you would do in wp-admin, by asking. Draft posts, fix SEO, clean out junk media, translate, pull your traffic numbers. You talk, it does the clicking.

Can I build a whole site with it from scratch?

More than you would think. Point an assistant at a clean install and it will scaffold pages, menus, images and content. It will not replace a developer on a big custom build, but it goes far. We run live sites built almost entirely this way.

Plugins

Is there an official WordPress way to do MCP?

Yes. WordPress 6.9 shipped the Abilities API, 7.0 built on it, and the official MCP Adapter exposes it over MCP. It is developer-grade plumbing; plugins wrap it with logins and a real interface.

So which plugin should I use?

If you are a developer, the official Adapter is the cleanest base. If you want MCP plus a chatbot and image generation in one place, AI Engine (ours) is built for exactly that. Either way, read the next answer first.

How do I spot a junky AI plugin?

MCP is this year’s gold rush, and a lot of plugins are marketing with some code bolted on. A big tool count and a shiny page prove nothing. Look for a real changelog, fast support, and a team that obviously uses the thing. You are handing an AI the keys to your whole site. Trust beats hype, every time.

Connecting

Which AI tools can connect?

Claude, ChatGPT, Cursor, VS Code, Gemini, n8n, and the list keeps growing. Set it up once and they all work. That is the whole magic of a shared standard.

What’s the coolest thing to connect?

Mix WordPress with other MCP servers, like GitHub, Slack, Notion, or a browser through Playwright, and an assistant can turn a GitHub issue straight into a published post. Just stick to official servers; plenty of community ones ship with no security at all.

What’s my site’s MCP URL?

Depends on the plugin, each has its own. AI Engine’s, for example, is /wp-json/mcp/v1/http. Check your plugin’s MCP settings for the exact one.

OAuth or API key?

Chat apps like Claude and ChatGPT: OAuth, just sign in and approve. Dev tools and automations: a token. Any decent plugin does both.

It won’t connect. What’s wrong?

Nine times out of ten, it is not the plugin. It is a local-only URL, a host or firewall blocking the path, or a login that quietly failed. Go outside-in: is the site public, does the endpoint answer, does the auth pass? (On AI Engine, there is a full troubleshooting guide.)

Trust, Cost & Models

Is it actually safe?

It can be, if you respect what it is. You are giving an AI real power, so scope it: a limited role, read-only when you can, revoke anytime. Test on staging first. Treat it like a new hire, not a magic button.

Will it leak my data?

Whatever the AI reads goes to whoever runs it (Anthropic, OpenAI, and so on). The MCP server is just the pipe. Expose less, sleep better.

Does it cost anything?

MCP is free and open, and most servers are free or already inside a plugin you own. The only bill is the AI itself, your Claude or ChatGPT usage.

Does the AI model matter?

Not to MCP. It is model-agnostic. Claude, GPT, Gemini, even a local model, they all speak the same protocol.

Got one we missed? Ask us, and we will keep this fresh as MCP keeps moving. 😺