Your AI-built app works on localhost. Here is everything that has to happen before anyone else can use it.
The agent says done and the app runs on your laptop. Then you try to send someone a link and nothing works. This is the full list of what is missing, in order.
The eight steps
- The code lives in a repo. Hosts deploy from a repository, not from the folder on your computer. Put .env in .gitignore before the first commit.
- Something runs the app all the time. Your laptop sleeps, a server does not. The app has to listen on the port the host gives it (usually a PORT variable) and on 0.0.0.0, not only on localhost.
- A real database. Data saved in a local file or in the browser's localStorage does not follow the app to a server. You need a hosted database, and the tables have to be created there too, so run your migrations against it.
- Secrets set on the host. API keys go into the host's environment variables, not into the code. The production values are usually different from the ones on your laptop.
- A domain and HTTPS. Without HTTPS browsers show warnings and sign-in often breaks.
- Sign-in knows the new address. Google sign-in only works for the redirect URLs you listed. The localhost one does not count. Add the production one.
- Logs. When it breaks at 2am you want to read what the server said, not guess.
- A way back. Keep the previous deploy so you can roll back when a change breaks things.
An agent can do most of this if you tell it the target. The part it cannot do alone is creating accounts and approving access, which is why the process stalls so often.
Quick check before you share the link
- Open the live URL in a private window. If it asks you to sign in with a localhost address, step 6 is missing.
- Create a record, reload the page, then deploy again. If the record is gone, the data is not in a real database.
- Search the repo for your API keys. If you find them, rotate them and move them to environment variables.
Prompt: take this app live
Paste this into Claude Code, Cursor, Codex or any agent that can run commands.
Read https://portal.traceshield.app/agent.md and follow it. Then take this app live. Check that it listens on 0.0.0.0 and the PORT variable, add a Dockerfile if it has none, and move anything saved in local files or localStorage into the Postgres database with migrations that run when the app starts. Keep secrets out of git and set them as environment variables. Deploy it, read the logs if the first deploy fails, and tell me the live URL.
The first time, your agent opens a browser tab and you click Approve once.
What Traceshield does here, and what it doesn’t
Traceshield covers the repo, hosting, Postgres database, environment variables, Google sign-in, logs and rollback from that one prompt. Your own domain is part of the Pro plan. Check the main page for what is live today.
More guides
Updated 2026-10-27. Written by the Traceshield founder.