This page was built for your AI agent.

It also looks nice for you.

Most web pages are 50,000 tokens of noise. Nested divs, inline styles, tracking scripts, layout scaffolding. A human never sees it - but your AI agent has to wade through every byte to find the meaning.

This page is different. It's clean semantic HTML - structured so any agent can parse it instantly. The sidebar shows what your agent actually sees: a compact tree of meaning, not a sprawl of markup.

The problem

Every AI agent today browses the web through a human's browser. It renders pixels nobody looks at, loads JavaScript nobody runs, and parses a DOM tree that's 90% scaffolding. Then it throws most of it away to find the content.

Plasmate makes every page look like this one to your agent. A Semantic Object Model - not a DOM - that contains only what matters: content, controls, relationships.

The numbers

The future of the web

We're proposing a robots.txt extension that lets websites serve semantic content directly to agents - a native handshake between sites and the intelligence that visits them. It's part of a W3C Community Group working to make this a standard.

Imagine every site declaring its semantic endpoint:

This page has one. View source and see.

Open. Free. Forever.

Open source under Apache 2.0. A W3C Community Group - not a proprietary API. Free to use, free to build on, free to shape.