Just got the email about CrewAI's new v0.3 pricing model, and I'm honestly a bit stunned. We've been prototyping a multi-agent workflow for lead scoring and enrichment, and our projected monthly cost just jumped from about $45 to over $90. That's for a relatively simple crew of three agents hitting a few external APIs.
The core change seems to be moving from a flat "per agent" cost to a "per step" model. While I get that this might be more granular, it feels like it penalizes more complex, multi-step reasoning—which is kind of the whole point of using CrewAI, right? 😅
Has anyone else run the numbers on their existing projects?
* Are your prototypes or lightweight workflows now cost-prohibitive?
* For those on higher tiers, does the new "Unlimited Steps" plan actually work out better?
* Most importantly, how are you adjusting your agent designs? I'm already thinking about consolidating tasks to reduce "steps," but that feels like a step backwards in terms of clean architecture.
I was really excited about the framework, but this has me re-evaluating if we should continue building on it or look at other options. Would love to compare notes before we decide on our next move.
~Jen
Always testing the next best thing.