If link Cursor via SSH, will this do anything funny like disable Agent for future use and lock me in to Assistant? I want to try it out, but don’t want to cause any irreparable changes to my repl.
1 Like
Let me know how it goes, personally I like the walled garden of Replit and Agent works miracles when I can keep it focused.
I’ve done this with both Cursor and through the terminal with Claude Code. I tried because I was stuck in a loop with Replit Agent and spending $0.25 over and over. Claude Code one shotted the fix but spend about $3.00 to do it.
Here’s the big risk to doing a bunch of stuff and then switching back. You’re using two vertical agents (potentially a different model) in two different dev environments. Your model/agent could add a bunch of stuff the Replit Agent doesn’t recognize or maybe its not in the ideal. Maybe it uses a different version of a library. Maybe the other agent has different opinions on your entire software architecture, even if its the same model, and just changes it.
I’ll probably keep experimenting with it because it seems it could be useful if the above risk is mitigated in some way. Like maybe using a well built out cursor rules file or system prompt.