Hey {{first_name}}
WebMCP was the big buzz word with SEO and GEO and all the other acronyms in July when Shopify rolled it out.
But I was a bit reluctant to comment as there was and still is a lot of debate about the usefulness of it.
Right now, the usage of it is low… very low but that does not mean its going to stay that way.
Codex, ChatGPT Work, Claude Code, GrokBot these AI Agent tools are becoming more and more popular and people are doing more of their research and purchaisng
This week I spent time setting up WebMCP on an unpublished version of Blend's website.
The good news is, Shopify has already added ten standard WebMCP tools to every Liquid store.
So an AI agent that supports WebMCP can search your products, check variants, read policies, add things to the cart and head towards checkout.
Basically, Shopify has given it a map, a trolley and the till.
There was one small problem for…
We are a CRO agency.
Nobody needs help adding a CRO audit to a basket.
But that made something quite obvious.
Shopify can help an agent complete an order.
But it cannot automatically tell someone:
Which size will fit them.
Which part works with the thing they already own.
How much they actually need.
Or whether they are about to make an expensive mistake.
And that is the where these standard Shopify WebMCP tools falls short.
The agent can use your store.
It just can't help someone decide what to buy.
The great news is you can add more custom tools.
Tools that helps the agent:
find the right size.
Check whether two products work together.
Builds a complete kit.
Work out how much someone needs.
Estimate when they are likely to run out.
That is the stuff that helps people Buy Now, Buy More and Buy Again.
And all this knowledge probably already exists inside your business.
It is just hiding in product pages, clunky quizzes, spreadsheets, support replies or Sarah's brain because Sarah has worked there for nine years and somehow knows everything.
A custom WebMCP tool gives an AI agent a proper way to ask for that information and get an approved answer back.
I have added 12 custom tools to Blend's draft site.
They help an agent understand the problem a Shopify brand is facing, match it to our research and find the most relevant services, A/B tests, case studies, etc,
Some of it worked beautifully.
Google's inspector found all 12 tools and ran one.
Codex ran another.
But then one ChatGPT test found the tools and then just decided to do its own research not with the tools.
So yes, it works.
Sort of…
And I say sort of because this is not yet widely adopted and the standard… yet.
Right now, I would not rebuild your entire eCommerce operation around WebMCP.
It is still just a proposed standard, support and usage is patchy, and adding tools will not magically make ChatGPT find or recommend your brand.
The agent has to reach your website first.
But what I would do is start mapping the opportunity now.
You do not need 20 custom tools.
Start with the question your customers ask every day and your best team member can answer in their sleep.
Write down the rules, the source of the answer and the weird exceptions.
Then make that first tool read-only and test whether an agent gives an answer you agree with.
Even if WebMCP adoption crawls, you will still end up with clearer product pages, better quizzes and fewer useful IP trapped inside Sarah's head.
Chat soon,
Peter
P.S. If you need help building out your non-standard WebMCP tools, give us a shout.
|
|
||||
|
|||||
