AI Agents Can Now Open Accounts, Buy Domains, and Ship Code — No Human Required
Cloudflare and Stripe just launched a protocol that lets AI agents create accounts, purchase domains, and deploy live apps — all on their own. Here is what that actually means for you.
By Troy Brown
Up until this week, AI agents were mostly doing knowledge work. Researching, drafting, summarizing, organizing. Useful stuff, but still in the realm of words on a screen.
That just changed. On April 30, Cloudflare and Stripe quietly launched a new open protocol that lets AI agents do something they have never been able to do before: open accounts, buy things, and deploy real software to the internet. No human clicking through forms. No manual checkout.
An AI agent can now create a Cloudflare account, register a domain name, start a paid subscription, and push a live application to production — end to end, without a person touching a single button.
Read that again. We are not talking about an AI that writes code and then waits for someone to copy-paste it into a hosting dashboard. We are talking about software that handles the entire process from idea to live website, including the part where money changes hands.
Here is how it works. The protocol has three layers: Discovery, Authorization, and Payment. Discovery lets the agent browse a catalog of available services. Authorization verifies the user's identity so the agent can create or link accounts on their behalf. Payment handles the transaction through Stripe's tokenized system.
The important detail on the payment side: your credit card number is never shared with the agent. Stripe issues a payment token instead. The agent uses the token to complete a purchase, but it never sees or stores your actual card details. That is a meaningful design choice.
There is also a hard spending cap. By default, Stripe limits any agent to $100 per month per provider. Not a soft warning — a hard stop enforced at the API level. If the agent tries to spend more, the transaction gets rejected. You can configure this limit higher or lower, but the default keeps things from spiraling.
You can also restrict what your agent is allowed to buy. Projects can be locked to specific merchants, product categories, or services. An infrastructure agent might only be approved to transact with certain cloud providers. An IT agent might only have access to a defined software catalog. Every transaction gets logged with full attribution, so you can see exactly what the agent bought, when, and why.
This is all part of a bigger launch from Stripe called Projects, currently in open beta. Think of it as an operating layer that gives AI agents a structured way to discover, authenticate, and pay for services — without requiring a human to babysit every step.
Cloudflare is the first major provider to support the protocol, but the design is open. That means other cloud platforms, SaaS tools, and service providers can plug in too. If this catches on, your AI agent could eventually sign up for hosting, connect a database, configure a CDN, and go live — all from a single prompt.
For small business owners and creators, this is a big deal even if you never touch the protocol directly. It means the tools you use are moving toward a world where an AI assistant can handle not just the thinking, but the doing. Want to spin up a landing page for a new product idea? The agent could handle everything from domain registration to deployment.
Of course, this also raises real questions. How much autonomy should an agent have with your money? What happens if it buys the wrong thing? What if a bug causes it to provision services you did not want? The spending caps and merchant restrictions help, but we are clearly in early territory here.
The human is not fully out of the loop, to be clear. You still have to accept the terms of service. You still set the budget. You still decide what the agent is allowed to do. But between those guardrails, the agent runs on its own.
This is the kind of infrastructure shift that does not make front-page news but ends up mattering more than most product launches. Agents that can only talk are assistants. Agents that can act, spend, and build are something closer to employees — and that changes the economics of running a small operation.
The takeaway is simple: AI agents just got a wallet and a set of keys. The spending limits and audit trails are real, which is encouraging. But the bigger story is the direction. We are moving from AI that helps you think to AI that helps you ship — and that transition just got a lot more concrete.
Subscribe
Get the next issue in your inbox.
Join The AI Signal for clear weekly notes on tools, workflows, and the handful of AI developments that are actually worth your attention.