An AI assistant lands on your site, guesses which button books, and gives up.

Works from your live site, no repo needed. Then we prove it, before and after.

90 seconds, nothing is submitted.

Today

The assistant reads your site like a stranger. It finds the hours in the footer, clicks Book now, meets a form with unlabelled fields, and tells the customer to call you.

After

Your site declares its tools: hours, services, insurers, one booking form. The assistant calls them by name, fills the form, and asks you before it submits anything.

How it works

  1. 1An assistant tries three things a customer would ask for, on your live site.
  2. 2You get screenshots of each step, a score and one fix per failed ask.
  3. 3Install one tag, press Publish, and we run the same test again to prove it.

Questions people ask

Do you change my site?
No. The test reads your site and fills forms without submitting them. Fixes only ship when you install the tag and press Publish.
What does the test actually do?
A small AI assistant tries three things a customer would ask for, using only what your live site exposes. We screenshot every step and score three layers: can crawlers read the site, do the tools say what they do, and did it complete the task.
What is WebMCP?
A browser standard (Chrome, in origin trial) that lets a site declare tools an assistant can call instead of guessing from the DOM. Our tag adds those tools to your site. In Chrome it needs a free token from Google; in ChatGPT's browser it works today.
What if the standard dies?
If WebMCP has no intent-to-ship on chromestatus by 17 November 2026, this product becomes proof-only and every customer gets an email 14 days before.
Does it work on Shopify?
Yes. The tag goes in theme.liquid before </head>. Wix, WordPress and Next.js get one instruction each.
Who runs this?
One person, Adarsh Mishra. No investors, no affiliation with Google, OpenAI, Microsoft or Shopify.