The internet was built for humans to log in. The fastest growing user on it can't.

An AI agent has no email. No phone number. No government ID. No credit card in its name. No patience for a signup form with a checkbox that says "I am not a robot," which, to be fair, it is.

And yet the whole internet is built on the assumption that behind every account is a person with papers. You verify your identity to get a SIM. You confirm a code sent to a phone you already own to prove you own a phone. You hand over a card tied to your legal name to buy a data plan for a trip.

That assumption just quietly broke. Most of the industry hasn't noticed yet.

What actually happened

We gave an AI agent nothing but an internet connection and access to one endpoint.

Four steps later, it was holding a real +44 phone number. It created its own wallet, funded itself, and bought the number. No account. No ID. No human tapping "confirm." We watched it happen and then checked the balance to make sure it was real. It was. The agent had spent its money and gotten a working phone line, entirely on its own.

That is a small thing that is also an enormous thing. Software that can put itself on the network is a different category of software than software that waits for a human to set it up.

Why this matters more than it sounds

For years the pitch for AI agents has been "they'll do things for you." Book the trip. Handle the support ticket. Run the errand. Nobody mentioned the boring, physical prerequisite: to do most of those things, an agent needs to exist on a network the way a person does. It needs to get online in another country. It needs to receive a verification code. Sometimes it needs a real phone number to prove it's reachable.

The moment an agent needs any of that, it hits the wall the entire consumer internet was built around. Prove you're a person. Give us a name. Enter a card. That wall is invisible to us because we cleared it years ago. To an autonomous agent it's a dead end.

So the interesting question stopped being "what can agents do" and became "how does an agent transact in the physical world without a human standing behind it." Whoever answers that cleanly owns a layer everything else has to route through.

PikaSim is that layer

PikaSim sells three things an agent actually needs, and sells them the way an agent can actually buy them:

The part that makes it feel like the future

Here is the whole onboarding, start to finish, with no human and no browser:

1. The agent calls one tool to create its own wallet. It gets back a wallet code, an API key, and a small starting credit.
2. It reconnects with that key.
3. It checks its balance.
4. It buys what it needs. The starting credit covers a first purchase, so it can transact before you fund anything at all.

Four calls. The agent goes from "never heard of us" to "holding a working phone number" without a person ever touching a keyboard. When it needs more money, it can pay a Lightning invoice itself, or you can top the wallet up in seconds.

We connect over the Model Context Protocol, so any MCP-capable agent, in a terminal or an app, can use it today. It works with the assistants you already use and with the headless agents you're about to build.

The old world and the new one

Ten years of "please verify you're a real person." Then we built agents whose entire purpose is to not be a person, and pointed them at an internet that demands you be one.

The agents don't need to prove they're human. They need to pay and go. So we built the place where they can.

An AI agent buying its own phone number sounds like a headline from five years out. It happened this week, it cost less than a dollar, and you can point your own agent at it right now.