In design · building now

Someone already built it. Get it running in two minutes.

Latropia is an app store for software made with AI. Find a tool by its screenshot, press Get, and it deploys to your cloud or your own laptop — no terminal, no engineer. Every line of source is right there, scanned and reviewed, if and when you want to look.

No spam. One email when there's something to try.

The problem

AI made writing software free. It did nothing for distributing it.

The tools people generate are genuinely useful. They just never reach anybody else.

For the person who made it
The good stuff dies on laptops
Someone builds something useful in an afternoon. It helps exactly one person, forever, because hosting and configuring it still takes an engineer.
For their company
Nobody knows what's been built
Every knowledge worker now ships software. Almost none of it is inventoried or reviewed. It runs on personal laptops, on personal cloud accounts, with company data in it.
For everyone else
You can't safely reuse it
Even inside one company, “Sarah built something that does that” ends in a direct message and a zip file. No way to know what it does or whether it's safe.

How it works

Four questions in plain language. Then it's yours.

Every app declares what it needs in a manifest an LLM writes when it's published — so a configuration form nobody understands becomes four questions anybody can answer.

Where should it run?
Cloudflare, Vercel, or your own machine. We're deliberately neutral about where it lands — including nowhere but your laptop.
Connect
One sign-in to the host you picked. Skipped entirely if you chose your own machine.
What it needs from you
If an app needs a login for another service, we'll make that clear and help you get set up.
It's yours
Storage is set up automatically — nothing to configure. Your sign-in details go to your host, never to us.

Why it's different

We run every app before we list it.

That one decision is where almost everything else comes from.

Proof of run
Screenshots that can't lie
Every submission is built and booted in a sandbox, and we capture the screenshots there ourselves. If it doesn't run for us, it doesn't get a deploy button.
Security
A catalog that tests itself
Static analysis, dependency and secret scanning on every release — and because we already boot every app, dynamic testing against a live instance. Findings come back as fixes, not a backlog.
Neutral by design
Your cloud, or none at all
Every other deploy button is a funnel into somebody's hosting. Ours goes wherever you already are — including nowhere but your laptop.

For agents

Ask your AI to find one and run it.

Latropia has two users, and one of them is an LLM. Every page has a machine-readable twin, and the whole catalog is available over MCP — so your coding agent can compare candidates on what they actually do, how well they're trusted, what they cost, and what they'll need from you. Then it just deploys the right one.

In practice
“find me the best expense tracker that imports spreadsheets and handles multiple currencies, and get it running”
→ Your agent compares four, checks they do what you asked, picks the one that's free and needs nothing from you, and deploys it. Nothing to approve. It just sends you the link.
You're asked to confirm only when it would cost money, need a sign-in of yours, or touch someone else's data. Everything else, your agent just does.

Where we are, honestly

Fully designed. Not yet built.

  • The product and architecture are worked out — including the packaging standard that makes one-click deployment possible across any host.
  • The interface is designed, down to a working click-through of the flow that takes you from a listing to a running app.
  • The catalog isn't live yet. That's what's being built now.

We'd rather say that plainly than dress up a waiting list as a product. Leave your email and you'll hear from us once when there's something real to try.

Who's building it

KF

Kevin Fealey. I've spent my career in application security. I co-founded FealzSecure, and AppSecAI, which automates security triage and generates code fixes.

Running other people's code safely is the hard part of this, and it's what I know.