WebMCP implementation WebMCP: your website, ready for AI browsers

Chrome, Edge and Comet already have an AI agent in the side panel, and that is only the beginning. Browsers are about to switch on WebMCP, and the agent will get things done on websites for the user. Get your website ready now: one script, no rebuild.

Let's talk about your website

/ they trusted us

The agent in the browser is only the beginning

An AI agent in the browser is already everyday reality: Chrome, Edge and Comet all have one. For now it reads pages and clicks on the user's behalf, but the next step is on its way. WebMCP lets a website tell the agent what it can do. Chrome and Edge are testing exactly that, and the biggest brands are already preparing. Get your website ready now and you'll be there on the day browsers switch it on for everyone.

Today
An agent in the side panel

It reads pages, answers questions, handles simple tasks. On a website without WebMCP it guesses where to click, and often gives up.

Chrome, Edge, Comet
Soon
Browsers switch on WebMCP

The website tells the agent what it can do, and the agent does it. Chrome and Edge are already testing this with websites, and the first platforms have it in production.

WebMCP in Chrome and Edge
New habit
The user says what they want

No browsing menus, no clicking around on the off chance. They state a goal: buy, get a quote, book, check, and expect a result. The website where the agent gets it done wins.

Your website with WebMCP

This is what buying looks like today

example-store.com/shoes?colour=black&size=9&price=under-100

Shoes
Clothing
Accessories
Sale

Search: running shoes

Colour

Black
White
Blue

Size

Price

under £100
£100-£160
over £160

9 steps to the basket

Menu
Shoes
Colour
Size
Price
Compare
Product page
Availability
Basket

This is what buying will look like with an agent in the browser

example-store.com/shoes

What the website can do for the agent

search products
show details
check availability
add to basket

AI agent in the browser panel

Find black running shoes in size 9 under £100 and show me a list with the key benefits

→ search products: running shoes, black, size 9, price under £100
← 3 models available in size 9: Model A £99, Model B £89, Model C £95
→ show details: Model A, Model B, Model C
Three options in size 9: Model A, £99, the lightest, for tarmac. Model B, £89, best cushioning, ships tomorrow. Model C, £95, waterproof, for trails. Which one would you like?
Model B, add it to the basket
→ add to basket: Model B, black, size 9, 1 pair
← Basket: 1 item, £89
Added black Model B in size 9 for £89, shipping tomorrow. Go to checkout?

A simplified flow using a store as the example: the user talks to the agent in the browser's side panel, and the agent uses the actions the website exposes instead of clicking through filters and product pages. We build a set of actions like this for any website, whatever its engine or technology.

The same website, two scenarios

Without WebMCP the agent looks at your website the way a person would, and guesses. With WebMCP it gets a list of actions from the website and runs them using your code. For the user it's the difference between a task done on the spot and looking elsewhere, at your competitor.

Without WebMCP

The agent analyses screenshots and page code to guess what is a button
It clicks on the off chance and gets lost in menus, filters and pop-ups
It answers in generalities or with data from other sources, sometimes wrong
It abandons the task halfway, and the user ends up wherever the agent managed to cope

With WebMCP by WebProfessor

The website tells the agent outright: "I can search, check availability, take an enquiry"
The agent runs actions with your code, in the user's account and basket
It answers with your data: offer, prices, terms, availability
The task ends with you: an order, a booking or an enquiry waiting for approval

The same conversation on every kind of website

The agent gets done whatever the user came for, whatever kind of website it is: on a company website it learns the offer, finds a service and sends an enquiry for a quote; in a store it finds a product and buys it; in an app it runs an operation or books a slot.

Company and service websites

The user learns the offer and gets in touch without reading the whole website.
- Understands the offer in a minute: the agent sums up what you do and for whom
- Finds a specific service and its scope
- Sends a quote request through your form
- Books a meeting in a free slot
Example request

"I need a new website for a haulage company with a quote form for jobs. Check whether they do projects like that, send an enquiry with my details and book a call next week"
Online stores

The user buys by talking, not by clicking through categories and filters.
- Searches for a product by features, not by name
- Compares variants, prices and availability
- Adds to the basket and heads to checkout
- Checks order status and the returns policy
Example request

"Find black running shoes in size 9 under £100 and show me a list with the key benefits"
Apps, SaaS, bookings

The user runs operations on their account with a single request.
- Checks free slots and books one
- Creates a ticket, an order or a report on their account
- Prices a chosen configuration
- Changes settings once confirmed
Example request

"Book the next free slot on Friday after 3pm and add it to my calendar"

What you gain when your website has WebMCP

Five things that change for your business once an agent can act on your website, not just look at it.

01.

You sell where the user ends the conversation with the agent

More and more decisions start with a question to an agent, not with typing in a web address. When the agent can find a service, compare products, add to the basket or book a meeting on your website, your website is where that conversation ends.

The result? Orders and bookings from a channel that, for most websites, does not exist yet.

02.

Contact without filling in a form

The user tells the agent what they need, and the agent carries it into your form. One click to approve, and the enquiry lands where it does today.

The result? Fewer abandoned forms and more complete enquiries, because the agent asks about the things the user would never have typed in.

03.

The agent answers the repetitive questions for you

Scope of a service, availability of a slot, delivery terms, opening hours: the agent reads it off your website and answers straight away, at any hour, with no waiting for an email.

The result? Fewer messages and calls asking things that are already on the website.

04.

You see what users ask the agent on your website

Every action the agent triggers on your website gets logged: whether it searched for a service, checked a slot or filled in a form, on which of your pages, and whether from the browser or from an assistant outside it. Your website only, no personal data.

The result? You know what users look for on your website and what they can't find, which click analytics will never show you.

05.

First in your industry

Few websites have any layer for agents today. Whoever adds one now gathers experience, data and traffic before the rest of the market starts catching up.

The result? A position in a new channel that you won't be able to buy with ads later.

Key dates: AI agents in browsers and WebMCP

09.2025
Gemini in Chrome for users in the US
01.2026
Persistent Gemini panel in Chrome and auto browse: the agent carries out tasks on websites
05.2026
WebMCP in Chrome's trial programme. Agentic Copilot in Edge for businesses
08.2026
The first large platforms ship WebMCP in production and in developer builds
11.2026
Planned end of the Chrome trial; the next step is availability for everyone

We make it happen with our WebMCP service

WebProfessor's WebMCP service gets your existing website ready for agents in the browser and for assistants outside it. You add one script, and we provide an index of your content, actions for agents, your form as an action, and maintenance on Cloudflare infrastructure.

One script on your website

Like Google Tag Manager. No change of hosting, content system or website code. In a browser without AI nothing changes.
An index of your content

Offer, products, articles and answers to questions, organised so an agent reads them without errors. Refreshed after changes on the website.
Actions for the agent in the browser

Company information, search, offer, reading content. A store or an app gets its own: basket, availability, booking, once connected to your system.
The same data for ChatGPT and Claude

Assistants outside the browser use the same source, so they answer questions about your company with your data.
Your form as an action

Your current contact form: the agent fills it in, the user sends it. Enquiries land where they do today.
Maintenance and keeping up with change

The standard evolves and browsers change the details. We keep your implementation in line with them, with no effort on your side.