Hey folks, quick sanity check. My headi is spinning with Ai Vs Agents Vs SSH … I am a product design lead, not a coder!
I’m building on Replit, but I’m getting tired of Replit Agent costs and the “agent did a bunch of extra stuff I didn’t ask for” problem.
I just connected Cursor to my Repl via SSH, and it feels like I can:
-
Do all coding and refactors in Cursor (with Cursor’s agent/models)
-
Use Replit mainly for hosting, secrets, database add-ons, deploy
-
Avoid using Replit Agent almost entirely
So… is Cursor too good to be true here?
Questions:
-
What are the real gotchas using Cursor + Replit over SSH (sleeping repls, ports, dev server, file watchers, latency)?
-
Does this actually reduce Replit AI usage/costs in practice, as long as I stop using Replit Agent?
-
Any sharp edges with DB migrations, environment variables, or long-running processes when working this way?
-
What would make you not recommend this setup?