Ah cool -- we weren't familiar with OpenAdapt. Will check it out.

One big decision we made was to focus on browser automations (instead of computer automation like Adept or OpenAdapt). The reason for this was that we wanted to leverage the information available inside of a DOM to improve the quality of our agent's actions. We found that relying on image-only analysis with X,Y coordinate interactions wasn't able to offer high enough reliability for production workflows