This is a great call-out. It's something currently in our roadmap Re: cost for execution. This really depends on the page, but currently it costs between 5 cen

This is a great call-out. It's something currently in our roadmap Re: cost for execution. This really depends on the page, but currently it costs between 5 cents and 20 cents per page to execute (today). We have an improvement planned to help it "remember" or "cache" actions it's done in the past so it can just replay them and bring the cost down to near zero. Re: LLMs it's capable of working with, currently it's only GPT-4V. I'll get this updated soon!