WebMCP: The Next Step in Building Websites for AI Agents
AI has moved from reading websites to using them. WebMCP is a proposed standard that lets a website expose structured tools an AI agent can use, to search a service, request a quote, or book an appointment, while the business still controls what those tools are allowed to do.
For years, websites have been built for two audiences: people and search engines. People need to navigate the site, understand what a business offers, and take an action. Search engines need structure, content, links, and metadata to understand what the site is about.
AI has already started changing that. We now optimise sites so AI systems can understand a business, which is what we mean by AEO, or Answer Engine Optimisation. But another shift is beginning: AI is moving from simply reading websites to using them, and that's where WebMCP gets interesting.
What is WebMCP?
WebMCP, or Web Model Context Protocol, is a proposed web standard that lets websites expose structured tools an AI agent can use. Instead of an agent trying to work out how a site operates by looking at buttons, forms, and menus, the site can explicitly tell it what actions are available.
The important part is that this isn't unrestricted access. The developer defines the tools and exactly what each one is allowed to do. Google describes WebMCP as a way for sites to expose structured tools so agents can interact with page features reliably, rather than simulating a person clicking around.
- Search for a service
- Request a quote
- Book an appointment
- Search and filter products
- Add something to a cart or begin a checkout
- Submit an enquiry
Why AI agents struggle with today's websites
Most websites weren't designed with AI agents in mind. They're visual interfaces built for people. A human sees a Book a Consultation button and immediately understands what it does. An agent has traditionally had to interpret the page, inspect its structure, and work out how to interact with it.
Say someone tells an agent to find an electrician who services Lake Macquarie, handles switchboard upgrades, and lets them request a quote. The agent has to find the business, open the site, confirm the service area, find the quote form, understand every field, and prepare the enquiry. Every extra step of interpretation is another chance for something to go wrong.
WebMCP takes a different approach. Instead of making the agent figure it all out from the interface, the site can expose something like a requestQuote tool, with structured information describing exactly what that action needs. The agent no longer has to guess.
Think of it as an AI-friendly control layer
A traditional website runs one path: a person uses the website, which fills a form, which talks to the backend. A WebMCP-enabled site keeps that human path and adds a second one alongside it, where an AI agent calls a WebMCP tool that triggers an existing website function that talks to the same backend.
We're not replacing the website. People will still want to see businesses, products, photos, reviews, pricing, and branding. WebMCP simply gives AI another structured way to interact with the same site, so it can serve both audiences at once.
WebMCP isn't the same as MCP
The names make this more confusing than it needs to be. MCP, or Model Context Protocol, is commonly used to connect AI systems to external tools and data, like a CRM, a database, or a business application through an MCP server.
WebMCP solves a different problem: it deals with the capabilities of the website currently running in the browser. A simple way to hold the two apart is that MCP connects AI to systems, while WebMCP helps AI interact with websites. They aren't competitors, and a sophisticated setup could eventually use both.
Declarative and imperative WebMCP
There are two main approaches. Declarative WebMCP is the interesting one for ordinary business sites, because an existing HTML form can be annotated so browsers and agents understand what it does. The visible form still works for customers, but an agent gets extra structured information about the form's purpose and its fields. It's progressive enhancement, not a rebuild.
Imperative WebMCP is for more sophisticated sites and web apps, which can define tools in JavaScript. Tools can even change with what's happening on the page, so a product-filtering tool might only appear once search results exist. The browser becomes the bridge between the website and the agent.
- searchServices
- checkAvailability
- filterProducts
- getQuote
- bookConsultation
- listAvailableAppointments
What WebMCP could mean for a local business
This is where it stops being a developer experiment. Imagine saying: find a carpet cleaner near me who can clean a three-bedroom house next Friday and get me a quote. Or find a restaurant around Newcastle with a table for four on Saturday night. An agent could discover suitable businesses and then interact with their sites on the customer's behalf.
Now picture two competing websites. Website A makes the agent visually interpret pages, find buttons, and work through a complicated form. Website B hands the agent a structured requestQuote tool. Which one gives the cleaner path to completing the customer's request? That's why WebMCP is worth paying attention to.
Discovery, then action: where AEO gets bigger
Getting recommended by an AI system and letting an AI agent complete an action are two separate problems. The first is discovery, the second is action. AEO focuses heavily on discovery: can an AI understand who the business is, what it does, where it operates, and whether it belongs in an answer?
WebMCP addresses what happens after the recommendation. You can think of the journey as understand, recommend, visit, act. Traditional SEO concentrated on the visit. AEO expands that into being understood and recommended. The agentic web adds one more question: can the AI actually do something once it arrives?
Do you need WebMCP today? Not yet, but build for agent readiness
No small business needs an expensive WebMCP rebuild right now. It's an early, proposed standard, browser and agent support is limited, and parts of the spec may still change. What does make sense is building sites that are ready for agents, which also happens to help people and search engines today.
There are security considerations too. Searching a catalogue is very different from buying a $5,000 product, and checking availability is different from confirming a booking. WebMCP is being designed with those distinctions in mind, including read-only operations and keeping a human involved where it matters. Good authentication, permissions, and validation matter more in this world, not less.
- Clean, semantic HTML
- Proper accessibility
- Logical, well-labelled forms
- Structured data and clear service information
- Sensible information architecture and well-defined actions
Pages worth acting on
Related page
AEO Packages
See how answer-ready content and AI search visibility are built into a site so AI systems understand and recommend the business.
View pageRelated page
AEO Central Coast
How AEO helps Central Coast businesses get understood and recommended by AI search and assistants.
View pageRelated page
Websites Built for Enquiries
The modern, well-structured build approach that also makes a site easier for AI systems to understand and use.
View pageRelated page
Talk About an Agent-Ready Website
Ask about building or improving a site with clean structure, structured data, and a solid foundation for the agentic web.
View pageWhat to read next
What's AEO and Why Does It Matter for Local Businesses?
A plain-English explanation of Answer Engine Optimisation for local businesses that want their websites to be clearer, more useful, and better positioned for AI-assisted search.
Read articleHow to Get Found in ChatGPT and Google AI Overviews
More local buyers are starting their search in AI tools rather than Google. The businesses that get recommended aren't necessarily the biggest. They're the clearest.
Read articleAI Websites vs Normal Websites: What Actually Changes?
A practical guide to the difference between a normal business website and one that's built to support AI tools, automation, cleaner lead handling, and future search changes.
Read articleQuestions local business owners usually ask next
Build a website that's ready for the agentic web
You don't need a WebMCP rebuild today. You do need a site with clean structure, structured data, and clear actions, so people, search engines, and AI systems can all understand and use it.