How to add a login to an AI-generated app

Short answer: The safest way to add a login to an AI-generated app is not to have the app contain a login at all. Deploy it to a platform where sign-on sits in front of every app — like Croft — so the app never handles passwords, and you just invite the people who should get in.

Login and access control are the riskiest code in any app. An AI can write you a login screen, but a botched one is exactly how data leaks — a missed check, a token that isn’t verified, a page that renders before it should. You don’t want that logic living inside an app your assistant generated in a few minutes.

The safe way

1. Deploy the app to your croft. Connect an assistant and ask it to deploy the app. How to connect →
2. You don’t build the login. Every Croft app automatically lives behind your croft’s enterprise single sign-on. Identity is checked at a gateway before a request ever reaches the app — so the app never touches passwords or credentials. It can’t ship a broken login page, because it doesn’t contain one.
3. Invite your people. Add them once by email and pick which apps each can open. They sign in with a passkey — nothing to remember — and every app they’re allowed to use already knows who they are.

Why this is better than AI-written auth

Because there’s no security code in the app for anyone to get wrong, and access is managed in one place. Someone leaves? Remove them once and they’re out of everything. Need finer control? Roles decide who can build, who can approve data access, and who can only use what’s shared.

Put simply: let the AI build the app, and let the platform handle the login.

More from the blog