Production

    Taking a Replit app to production.

    Replit is a fast way to get something running, and a slow way to keep it running. The problems show up later and all at once: the database was fine for a prototype and is not fine for a business, secrets sit in environment variables nobody has audited, there is no deployment pipeline because there was never a deployment, and the whole thing lives on an account belonging to one person.

    What moving off actually involves

    A real managed database

    Whatever the prototype used becomes managed PostgreSQL with migrations, backups and a restore you have tested at least once.

    Secrets handled properly

    Out of the repository and out of ad-hoc environment variables, into a secret store, with anything previously exposed rotated.

    Infrastructure in your accounts

    Vercel and a managed Postgres for most products, AWS with Terraform where the workload justifies it. Either way the account is yours.

    A real deployment pipeline

    GitHub Actions building, testing and shipping on every push, so releasing is not one person clicking through a hosting UI.

    Monitoring and real backups

    You find out something is broken from an alert rather than from a customer, and there is a way back when it is.

    The code in your repository

    With a real history, a README that works on a clean machine, and a runbook for the deploy that someone else has followed.

    The order we do it in

    01

    Read the code

    The actual code and whatever is standing in for infrastructure, not the demo.

    02

    Stabilize it first

    Anything actively losing you data or leaking access gets fixed before anything moves.

    03

    Lift and shift

    Database, secrets and hosting moved onto infrastructure you own, with the app still working.

    04

    Automate deploys

    Pipeline, environments and migrations, so the next deploy is a boring non-event.

    05

    Hand it over

    Documentation, a runbook, and a walkthrough with whoever will look after it.

    Answered before you ask

    Can the app stay online during the move?

    Normally yes. We stand the new environment up alongside the existing one, migrate a copy of the data, prove the replacement works, then cut over. Downtime is usually the length of a final data sync rather than the length of the project itself.

    Do you have to rewrite the whole thing?

    Rarely. Replit projects are usually ordinary Node, Python or React code, and the problem is where it runs rather than what it is. We rewrite when the data model would have to change anyway, and we tell you which of the two situations you are in before starting.

    What does it cost to move off Replit?

    The audit is small and can be bought on its own. The migration depends on how much infrastructure has to be built from nothing, which is what discovery is for. You leave that call with one price for the move, written down, rather than a range that shifts once the work has started.

    The person who built it has left. Is that a problem?

    It is a normal starting point rather than a problem. We work from the code rather than from a handover, and part of what you get back is the documentation that should have existed in the first place, so the next person is not stuck where you are.

    Related

    Tell us what it has to do.

    Describe what you’re building, what it has to do and when it needs to be live. You’ll get back the architecture we’d use, the trade-offs, and a realistic timeline, before anything is signed.