Cost Effective Usage of Cursor

Discussion
1

I’ve been using Cursor for almost 9 months now, intensely in the past 3 months.

Currently I’m seeing a bill of $20+ per day as we use it on a lot of projects and I’d like to know people’s views on how to make this more cost effective.

Reducing usage generally is not a possibility as we’re smashing through projects.

  1. Reduce use of MAX models to certain scenarios? If so, which scenarios and does changing the model as part of a flow lose info / context anywhere?

  2. Using our own keys for Anthropic, Google etc. ? We already have these for other purposes

  3. Something else we’ve not thought of

Info much appreciated and likely of use to others :raising_hands:t2:

1 Like
2

Yeah its tough and something I have been thinking about also. In a larger org with widespread usage, its tough.

I think education is important - without it I have seen some that will just keep MAX mode on the latest model turned on all the time, for simple tasks even, not realizing that it does not even pull from the alloted credits.

Encouraging people to experiment through their day to day to try different models and learn what works better with which / what prompting techniques are better with which etc…

But barring cutting off max mode access, I am not sure what the answer is yet - as there are some cases where it probably makes a valid difference.

So many variables also - greenfield vs brownfield, front end vs back end, small tools and 1 off small libraries etc to larger more complex multiple layered features that need more context etc…

But if the cost is negated by the rate of implementation - $20 a day vs a single hour of an engineers salary should be accounted for at the cost of stifling velocity and innovation…