Presstack: The First WordPress WebMCP Explained

What a WebMCP Actually Is

The Model Context Protocol (MCP) is an open standard that lets AI agents query a data source in a structured, two-way way. Think of it less like a sitemap and more like an API your site exposes specifically for AI. Where a sitemap tells a crawler which pages exist, an MCP endpoint lets an agent ask a real question and get a real answer back.

A WordPress WebMCP is that endpoint sitting on your WordPress install. It is not a rebranded RSS feed or a schema snippet. It is a live surface an AI assistant can call at query time, receive structured data from, and act on. Until now, nothing like this existed as a ready-to-install WordPress solution.

Why Most WordPress Sites Are Invisible to AI Agents

Traditional crawlability and agent-readability are two different things. Google's crawler reads HTML and indexes text. An AI assistant trying to answer a user's question operates differently. It needs structured, queryable context, not a page to parse.

A concrete example

Picture a user asking an AI assistant, "Does this plumber cover my area?" The assistant goes looking. The plumber's WordPress site has a coverage page, but it is plain HTML with a paragraph of text and an embedded image of a map. The agent has no surface to query. It either guesses from the text, returns a vague answer, or tells the user it cannot confirm.

That is the gap. The content exists, but there is no machine-readable endpoint the agent can call to get a confident, structured answer. The site is crawlable but not agent-readable.

Why this matters now

AI assistants are increasingly the first touchpoint between a business and a potential customer. If your site cannot answer an agent's query directly, you are simply not in the conversation. The lead goes elsewhere.

How Presstack's Agent Surface Works

Presstack ships three components that together form the agent surface. Each one serves a distinct purpose.

• llms.txt and llms-full.txt, plain-text read files that give AI language models a concise summary of the site's content and context. These follow the emerging llms.txt convention and are the first thing a well-behaved agent checks before querying further.
• /.well-known/agent.json, a manifest file that declares the site's agent capabilities, what the MCP endpoint accepts, what actions are available, and how to authenticate. Agents use this to understand what they can do before they do it.
• The live MCP endpoint, the queryable surface itself. An agent sends a structured request and receives structured data back, business details, services, locations, availability, and supported actions like submitting an enquiry.

No manual configuration is needed. Installing Presstack activates all three automatically.

What Agents Can Actually Do on Your Site

This is where it gets practical. An AI agent with access to a live MCP endpoint is not just reading your content. It is able to act on it.

Specifically, an agent can read your core business information (name, contact details, opening hours), check which services you offer and which locations you cover, confirm whether a particular service is available in a given area, and submit an enquiry or booking request on a user's behalf. All of this happens without a human visiting the site or filling in a form.

That changes the shape of how leads arrive. A user asks an AI assistant a question at 11pm. The assistant queries your MCP endpoint, confirms you cover the right area, and submits an enquiry. By morning, it is in your inbox. The site did the work.

Schema and Entity Linking Underneath It All

Raw data is only as useful as the context around it. Presstack's unified schema graph connects your business entities to recognised external references, including Wikipedia and Wikidata, so the data an agent receives is not just structured, it is trustworthy and unambiguous.

When an agent queries your MCP endpoint, it does not get a pile of HTML to interpret. It gets a clean, entity-linked data response. "Plumber" maps to a defined service entity. Your coverage area maps to geographic entities an agent already understands. That context is what separates a useful answer from an uncertain one.

Presstack handles the full schema graph automatically, including service-to-location mapping and P165 node triangulation for precise geographic coverage. You can read more about how Presstack's structured data approach works if you want the technical detail before committing.

Getting This Running on Your WordPress Site

The practical path is straightforward. Presstack is available as a full install or a lighter version for existing WordPress builds. Both include the WebMCP surface, the agent manifest, the llms.txt files, and the full schema graph.

There is no separate MCP plugin to configure and no schema markup to write by hand. Activation deploys the agent endpoint with the data Presstack has already mapped from your site's content and settings.

If you are running an existing WordPress build and want to add the agent surface without rebuilding anything, that is exactly what Presstack Lite is designed for. One install. The agent surface goes live from that point.

Is your site currently answering the questions AI assistants are already asking about your business?

Related: From Internal Build to Public Release: The Presstack Pre-Launch Roadmap