It reads
Paste the prompt below into the AI app you already use. It reads a plain-text copy of every page that I keep for exactly this, and answers from that.
- ChatGPT
- Claude
- Gemini
- Grok
Any AI can read this site. In Chrome, it can do more than read: every page hands it a set of tools, so it asks the site instead of guessing at it. That is WebMCP, and this site is built for it.
Checking…
Both work. Nothing here needs setting up before your AI can answer you. The second way is just better.
Paste the prompt below into the AI app you already use. It reads a plain-text copy of every page that I keep for exactly this, and answers from that.
Your AI works inside the page: it searches, opens posts, asks a figure what it means and runs the demos, through tools the page gives it. Faster, and it cannot misread what is not there.
Honestly: this is new. Few AIs use tools like this yet, and Gemini in Chrome may still need the setting below.
Only if the check above says so. Two minutes, once.
Menu → Help → About Google Chrome updates it, and says the version.
Chrome does not let a website open its settings pages, so copy this into the address bar:
chrome://flags/#enable-webmcp-testingSet it to Enabled.
The check at the top will say you are set.
Copy it into your AI. It uses the tools if it can, reads the text if it cannot, and answers you either way.
I'd like your help with Hetansh Mehta's website, https://hetanshmehta.com. I want to know who Hetansh is and what he writes about. If you can use a web browser: open the site there. Its pages give AI agents tools through WebMCP (document.modelContext). Call check_webmcp first, then use the site's tools instead of reading the page. If you can't use a browser or its tools: read https://hetanshmehta.com/llms.txt and the plain-text pages it lists. Either way, answer me first. Then, only if you weren't able to use the site's tools, tell me in one line how I could turn WebMCP on for a better experience.
WebMCP is a proposed web standard from the Chrome team: a page declares tools with names, descriptions and typed inputs, and an AI in the browser calls them, while you watch. Chrome’s documentation.
Where Chrome does not have it yet, this site brings its own copy, so the tools are there in any browser for an AI that knows to look. Without a browser, the same answers are at /llms.txt and /tools.json.
I wrote about what changes when a page hands an agent tools: Two doors.
Chrome runs WebMCP here as a trial until 16 Nov 2026. After that, this page and the check will say what changed.