$ systemctl status yourappBuilt with Replit
Your Replit app has a backend. That changes what breaks.
Replit Agent is the only builder in this group that regularly produces a full application: an Express server and its own database, not just a front end talking to somebody else's API.
Send the address, and say whether anything of yours has to keep running — a server, a job, a queue. That answer changes the work more than the code does.
What we measured on Replit
A server of its own — and nothing that restarts it.
ship a real backend — a server directory, an API, or Drizzle. 46 of them use Drizzle.
100 confirmed Replit Agent repositories, read 2026-08-03
have at least one source file over 20 KB. The median largest file is 28 KB — around 800 lines.
55 of those 100 that contain JavaScript or TypeScript
have a single test of any kind.
100 confirmed Replit Agent repositories, read 2026-08-03
Straight about what we know
What is true
Among the Replit projects built on Vite, 75% have no rewrite rule — the same routing failure as Lovable. But the backend is the part that matters here: a process that nobody arranged to restart will not come back after a reboot, and nothing in these repositories arranges it.
What we won’t claim
Getting to a clean sample was expensive and we would rather say so: 1,374 repositories had to be checked to confirm 100. The obvious marker, a .replit file, is present on anything ever hosted there — tutorials, forks, abandoned templates. Only replit.md means the Agent wrote it.
Read the measurements and judge how much of it looks like your app.
What breaks first on Replit
The first is shared with every Vite app. The rest belong to having a server of your own.
- 01
“It runs until the server restarts, then it's just gone.”
- 02
“Everything worked on Replit and nothing works on the VPS.”
- 03
“The database connection points at localhost in production.”
- 04
“One file is two thousand lines and I'm afraid to touch it.”
Every job runs the same way: we look from outside first, work on our own machine, and the result goes onto your infrastructure with you watching. How we work sets out what access is needed at each stage and what we never do.