Your content has one origin and many readers. A browser renders it for a person. A search crawler indexes it. An AI agent retrieves passages from it. A command line tool or an MCP client pulls it as data. Each reader gets a different view of the same source.
The mistake is to build for the browser and assume the rest see the same thing. They do not. A page that looks finished in Chrome can be an empty shell to a crawler that does not run JavaScript, and a wall of undifferentiated text to a model trying to find one fact.
Kenovar shows you all of these views at once, bound to one graph of your site. The origin is the thing you own. The views are what each consumer receives. When they disagree, that gap is the problem, and it is usually invisible until you look through the right eyes.
The origin is not the HTML page
It is every surface a machine can consume, and each one is a place a consumer can find, judge, choose and call you. The HTML page is one. So is an RSS or product feed, a structured-data block, an MCP endpoint an agent calls to run your tools, an A2A Agent Card an agent reads to discover what you offer and the files under /.well-known/. The Origin surface tool probes all of them and reports which exist, because a site that is perfect for a browser can be invisible to an agent that never loads a page.
This matters most where the page is not even involved. In agentic commerce, merchants push a product feed to the engine rather than have it crawled, so the feed, not the page, is the origin the transaction reads. But origin-side access still decides outcomes at scale: when a large retailer blocked two crawlers in its robots.txt, hundreds of millions of its products vanished from an AI shopping surface. Whether an agent meets your page or your feed, it meets some machine surface of your origin, and all five stages apply to each.
This is the heart of it. The browser is not privileged. Neither is the agent. Serve the origin well, across every surface, and every reader gets what it needs.