The short video below shows me using WordPress Playground with WebMCP to add a hamburger menu to a website. It’s pretty cool: I can ask AI to work on the site, see the results in my browser, and keep shaping things through conversation.
Want to try it yourself? Open the latest ChatGPT desktop app, switch to Work mode, and open WordPress Playground in its built-in browser. Once WordPress has loaded, try asking:
Use the available site tools to build a colourful homepage for a local coffee shop.
Then watch it take shape and ask for changes. You can check which tools are available by clicking Site tools in the browser’s address bar.
You’ll need a supported model. OpenAI currently recommends GPT-5.6 Sol or GPT-5.6 Terra. Availability depends on rollout, and Site tools aren’t currently available in Enterprise or Edu workspaces. This WordPress Playground guide explains the setup in more detail.
One of the hardest things about WordPress for beginners is knowing where everything lives. You know what you want: a contact form, a different layout, a page for your business. Finding the right screen, setting or plugin can be the difficult bit.
WebMCP lets a website give an AI assistant a clear set of actions it can perform. The assistant can then use those actions to help you get things done.
A bit like giving it an instruction manual with working buttons.
You may have heard of MCP, which connects AI assistants to tools and services through a server. Those connections can work without you having the website open.
WebMCP makes tools available through the webpage you’re using. You and the assistant can work in the same browser session, with the site right there in front of you. Both approaches have their uses, and a website can support both. More on the difference.
WebMCP is really fast
When AI operates a website through screenshots and clicks, it has a lot to figure out. Where’s the menu? Which button does it need? Did that click work?
WebMCP gives it a more direct route. The website tells the assistant which actions are available and how to use them. That can mean fewer steps, less guesswork and faster, more reliable results. Google highlights these benefits in its WebMCP introduction.
That’s a speed advantage over AI clicking its way around a page. It doesn’t mean WebMCP is automatically faster than MCP for every task.
WebMCP is designed for you and the assistant to work within the same web interface. You can inspect the result, make your own changes and guide what happens next. WebMCP specification.
For WordPress, that could feel like:
“Build me a page for our next gig.”
You look at the result.
“Make the heading bigger, move the booking details higher up, and use this photo.”
Then you jump in and edit the wording yourself.
That feels like a natural way to build a website. You can make decisions as the page takes shape, mixing conversation with hands-on editing.
WordPress Playground already gives us a place to explore this. It runs WordPress in the browser, and its WebMCP integration lets compatible assistants work on the running site. It can also make tools registered by plugins available to them, as the Playground team explains.
That’s where WordPress’s plugin ecosystem becomes really interesting.
A forms plugin could offer an action to create a form. An events plugin could help prepare an event. A shop plugin could make it easier to create a product. Developers could make the features people already use accessible through conversation.
That support needs to be built. WebMCP doesn’t automatically make every plugin AI-ready, and it’s still a draft proposal with developing browser support.
But the opportunity is huge: beginners could spend more time shaping their website and less time learning where everything lives.
Here’s a look at the finished page. I’m scrolling through the full design so you can see what AI created in WordPress Playground using WebMCP.
And, as the video shows, we can already start getting a feel for it today.
Leave a Reply