Why I built Open for Agents

I have been a web publisher and webmaster for more than 20 years. That is long enough to see a few big changes in how people use the internet.

We went from directories to search, from desktop to mobile, and from typing in an address to finding almost everything through a platform. I think the next change is already taking shape.

People will still visit websites, but they will do less of the routine browsing themselves. We will have personal assistants that can go out, find information, compare options, check what has changed, and carry out useful tasks for us. Those assistants will have to deal with websites every day.

Machines are already using our websites

As a publisher, I can already see how much website traffic is not a person sitting down to read a page.

Some of it is familiar search crawling. Some of it appears to be bots, automated systems, and AI-related crawlers. I cannot always tell whether a particular request is collecting search data, training a model, looking for an answer, or simply probing the site.

What I can see is the behaviour. Automated traffic follows broken links, requests hundreds of variations of the same page, repeatedly lands on 404s, and consumes hosting and infrastructure without behaving anything like a human visitor.

Blocking every machine is not a realistic answer. Letting every machine guess its way around the site is not a good answer either.

The missing interface

Most websites were built for people. The meaning of the site is spread across page copy, menus, forms, buttons, and visual layouts. A machine has to crawl all of that and infer what matters.

When I saw WebMCP, the direction made sense to me. A website can provide explicit tools with defined inputs and structured results. An agent does not have to pretend to be a person clicking around a page when the site can tell it, clearly, what is available.

But a protocol alone does not answer the question I care about as a website owner:

What should an agent be allowed to know and do on my site?

That is the reason I built Open for Agents.

Open for Agents shows the owner which site features could be offered to agents. The owner reviews each one, publishes only the ones they want to offer, and can check what an agent can reach.

I started with WordPress because it is the platform I know from years of real publishing work, including the unglamorous details that decide whether something actually works: plugins, caching, redirects, forms, shops, and public hosting.

WordPress, Shopify, and Any-Web

The same need appears in different forms, so Open for Agents now covers three related products:

• Open for Agents for WordPress lets an owner choose which site information and capabilities to publish for agents.
• Open for Agents for Shopify is a separate, public, read-only app that checks whether product information on supported AI shopping services still matches the merchant's store.
• Any-Web is a controlled, read-only preview where people can ask questions about selected public Open for Agents material and follow the source links.

The optional Open for Agents Assistant adds a conversation to a WordPress site. It uses the owner's reviewed capabilities, shows consequential actions before they happen, and keeps checkout, payment, and sensitive account work with the visitor.

Website owners should decide what agents may use

I am not trying to make every website autonomous, and I am not building against agents. I expect agents to become a normal way to use websites.

Agents can be genuinely useful. Leaving them to crawl and guess their way through a site can also create unnecessary cost and risk.

Publishers, merchants, and site owners should decide which information and actions agents can use. Agents should use that published interface instead of reconstructing one from scraped pages.

If you run WordPress, the best place to begin is the current Core plugin on WordPress.org. If you want to see the idea working first, try the live Storefront Assistant.