The customer sets the goal
A customer asks their assistant to find an appropriate service and an appointment next week. The assistant visits the business’s website.
From browsing to getting things done
Can an AI agent use your website?
A customer may ask an AI assistant to find a service, check an appointment or prepare an enquiry. WebMCP is a proposed standard that helps websites make those tasks clearer for AI agents. For your business, the opportunity is a smoother customer journey, backed by reliable information and sensible controls.
In plain English
What is WebMCP?
WebMCP stands for Web Model Context Protocol. It is a proposed web standard that lets a website describe specific actions an AI agent can use within the customer’s browser. Those actions might search for a service, check an available date or prepare a form. The agent gets a clearer description of what each action does and what information it needs.
It builds on the website people already use, with the customer involved in the journey. The proposal is still evolving and browser and agent support varies. Start by checking whether it would help a real customer task and testing in a supported environment.
Being found is one step. Helping someone complete their task is the next. WebMCP describes actions available on a page once an agent visits it; it does not guarantee higher search rankings or that an AI assistant will recommend your business.
Based on Chrome’s WebMCP documentation, which describes the proposal, current support and limitations.
Why it matters
Make the next step easier.
A helpful website lets customers do something useful: find the right service, check details and take the next step. An AI agent needs to understand those same tasks. WebMCP gives the website a way to describe them explicitly, reducing the need for the agent to interpret every button and field.
The business case is practical: fewer abandoned journeys, more complete enquiries and less time correcting missing information. These are outcomes to test and measure in a pilot, rather than benefits to assume.
An illustrative customer journey
“Help me book a suitable appointment.”
Clear actions, reliable answers
The website could describe actions to find services and check live availability. The assistant uses the customer’s requirements to return relevant options.
Review before commitment
The customer reviews the service, price and appointment. The business requires confirmation before a booking is made, then returns a clear result.
The website and booking system still enforce availability, permissions and business rules. WebMCP provides a way to describe and invoke the actions; the business must design and enforce the approval and security controls.
A practical readiness check
Before you make an action available, ask:
Is the task clear?
Can you describe the outcome, required information, decision points and exceptions without relying on someone’s unwritten knowledge?
Is the information dependable?
Are prices, availability, service details and customer data accurate, current and owned by someone?
Are permissions defined?
What can anyone see, what needs a signed in user and which actions require a person to approve them?
Can systems complete the work?
Will the website, forms and connected systems return a trustworthy result, including when something goes wrong?
Can you check the outcome?
Can you see what an agent requested, what happened and whether the customer received the right result?
Who owns improvement?
Is someone accountable for testing, maintaining and pausing the tools when they behave unexpectedly? Can they restore a dependable customer journey?
Start small
Choose one useful journey.
A good pilot could help someone find the right service or prepare a detailed support request. Define the task and its boundaries, make the underlying process dependable, then test it with real scenarios and human review.
Leave high impact actions such as payments or changes to customer records until permissions, confirmation and recovery are clear. The best first step may be fixing the process behind the page.
Use the Five Pillars of Successful Transformation to measure the pilot, with clear conditions for pausing or stopping.
Explore the wider AI Readiness Assessment →Project Reboot
Where should you begin?
I help organisations understand the customer journey, information, systems and rules behind their website. Together we can identify a useful first task, assess the foundations and define a pilot with clear measures of success. That includes deciding whether WebMCP is relevant now or whether improving the existing process should come first.
Learn about Steph’s experience or explore the case studies.
Talk about your readiness