prashar32

I'm trying to understand the use-case for this, because I have personally seen claude code generating this much code in one go, it totally depends on the prompt and the setup made up. Also code quality is more important than the lines of code. Planning the task first and in-between stages of work is correct to get the better quality and results but Claude code, Codex both does this already

No dedicated adapter yet for Crew AI, but it could be still governed yet, as CrewAI runs its LLM call through LiteLLM, so point the model at Rikkernel by setting the base URL, this could happen without any code changes.

If you are running CrewAI for real, happy to open an issue for it and scope it.

yes, its working you can try it out and let me know the feedback