Cloudflare 让 AI 代理付费访问，催生新商业模式

Cloudflare launched something huge around AI agents and I think a lot of people haven't really paid attention, but I think the people that pay attention are going to be able to create businesses that monetize in a completely new way that I think is really interesting. So, I wanted to do an episode breaking down everything. By the end of this episode, I want you to understand three things. The first is what is Cloudflare actually doing here in plain English around AI agents.

The second is why does this create a new business model for the internet? And the third is startup ideas. I think you should actually build on top of this. I want to show you the wedge. I want to show you who the customer is, what the first version looks like, how I would sell it, and how I would become a real company. The opportunity is way bigger than Cloudflare launched a thing here. The opportunity is agents are going to need clean, trusted, useful resources to do their jobs.

And the people that figure this out and launch things in the next 18 months are going to do incredible things. I'm going to show you exactly how everything works in complain English. I have no affiliation with Cloudflare. I just want to see you win. I'm rooting for you and I'll see you in there. Enjoy the episode. [music]

Let's start with the old bargain. That's the way I think about the old bargain of the internet. For a long time, if for a long time, if you owned a website, you let search engine crawl it because search engines sent you traffic. And then Google would read your site, Google would index your site, someone would search, your page would hopefully show up and then a human being would click through. Once the human being landed on your website, you would monetize that attention.

Maybe you show ads, maybe you'd capture their email, maybe you'd sell them a subscription or do some affiliate thing. That was the trade. So the crawler got the content, the website got the visitor, the visitor got the answer, and the model funded a massive part of the internet. Now AI agents are changing this flow because an AI agent or AI system could read your page, pull the useful answer, give that answer to the user, and the user might not visit your website.

So the content still created value, but the website may lose the visit. And if the website loses the visit, it loses the ad impression, the email capture, right? The affiliate click, all that stuff, it just loses it. So that's why a lot of publishers are upset. But I actually think the publisher conversation is only the first inning. The bigger idea is that agents are going to use the internet in a way software uses infrastructure.

So, uh, they'll request things, they'll call tools, they'll compare products, they'll retrieve data. By the way, I'm going to get into how CL Cloudfare plays into all this, uh, in a second. They'll buy access, uh, they'll take action, and that basically means that the internet needs a new pricing for machine usage. Obviously, a human being doesn't want to pay 0.003 cents to read a page. That would feel ridiculous. If I clicked on a recipe and it said, "Please pay 1/3 of a cent to see the sauce," I would close the laptop and probably order tacos out of spite.

But a machine doesn't care if the payment is tiny and automatic. If the data helps the agent complete the job, the agent, you know, can pay. And that's really the mental model. The human web monetized attention and the agent web monetizes useful resources. So what is Cloudflare actually doing here? Well, there's actually there's a few pieces I want to go over. The first is AI crawl control, which gives site owners visibility and control over AI crawlers.

You can see a crawler activity. You can allow certain crawlers to do things and you can block certain crawlers. you can understand who's accessing your content. They also announced pay per crawl and that's the direct monetization piece. That's the piece that's going to get people paid. A site owner now can charge AI crawlers when they access content. The crawler could either present payment intent in the request or receive a 402 payment required response with the pricing.

Then monetization gateway is even a bigger version of that. So, Cloudflare is basically saying this should not just apply to crawlers reading pages. Any resource behind Cloudflare could have payment rules, a web page, a data set, an API, you know, an MCP tool call, basically a premium endpoint, a file, a search index, and the payment rail they're talking about is X42, which uses the HTTP 402 payment required status code.

The agent requests the resource. The server says, "This costs this amount, usually a fraction of a penny, and the agent pays and retries with proof of payment." Cloudfare can verify at the edge before the request hits the origin. Basically, in plain English, CloudFire is trying to make paid access feel like part of the internet itself. So, you don't have this giant checkout flow. You don't have to create an account first.

There's no sales call. there's no like enterprise procurement stuff. Basically, there's a request that becomes uh the transaction. And that's a really really big idea. Um and it's a big idea because the request becomes the transaction and then there's these tiny resources on the internet that on the internet that could become businesses. And that's what we're going to get into. That's why I'm doing this episode. I'm doing this episode not just because cloud for launch is thing.

Um, it's interesting because of all the businesses that could be built on top of it. Uh, a data set could charge per lookup. An API can charge per successful call. A research archive could, you know, charge per answer. A product catalog can charge per comparison. You get the idea. And, you know, someone in the comment section is basically going to say something like, "Yeah, but you know, agents don't have wallets." Well, we're moving into a future where agents are going to have wallets.

They're going to have email addresses and assuming that's true and I think you know there's a good chance that that happen is happening happens because it's already starting to happen. They're going to make transactions and Cloudflare's AI index points in the same direction. So they've talked about making websites easier for AI systems to use things like MCPS, LLM.txt, txt, search APIs, and bulk data APIs. That matters because agents need access, but they also need structure.

So, humans can tolerate messy websites. It might not be a good experience, but they can tolerate it. Um, you know, we click around, we zoom in, we read the FAQ from 2019, we open a PDF. We're good at suffering human beings. But agents need really clean doors. So they need information in a format they could trust and use. So the shift is websites become resource layers, content becomes indexes, expertise be becomes scalable, data becomes metered, and tools become agent accessible.

Um, and that's what I'm paying attention to as a founder because I can build on top of it. So there's this stack that's forming uh around this whole new internet and I want to go through it. It's actually simpler than we think. So, first you have a messy internet. That is PDFs, pricing pages, old blog posts, you know, YouTube videos, support docs, um comparison sites, websites that look like they were designed uh when people still said web 2.0 without any irony.

Uh then someone cleans that into structured data. Then someone makes it agent readable through an API, MCP tool, search index, feed, LLM.TXT or some other clean access point, but it needs to be cleaned. Then someone adds payment rules. Some things are going to remain free because, you know, free creates a lot of distribution, but some things are going to be paid because they're extremely valuable and some things are going to be blocked because they should stay private.

Um, and then someone adds trust in analytics like is the data fresh? Is the s