Sometimes you just want to experiment with some agentic changes and then you end up stuck - it would be nice to have a localised version control system of changes that the AI makes within cursor. That way, you could roll back to a previous point, and try a different prompt before things go off the rails. From a user perpective, this could be represented graphically with a graph and the user prompt for each point in time. Maybe it’s just me though!
1 Like
Hey, you can already do this with checkpoints within the chat: