The protocols designed for Kitty are great, but I when I last gave Kitty itself a try, I felt it was extremely opinionated in a very off-putting way.

It was advertised as being highly extensible and scriptable, but despite that, I found it to be less flexible than my old setup which wired up tmux with Python scripts. For instance, I was able to track which pane was previously active when creating a new split, allowing me to have the newly-created pane's bashrc read the old pane's bash /proc/ entry to copy the environment variables. That wasn't possible in Kitty. And although Kitty's splits layout were functional, resizing the splits themselves was an unconfigurable pain in the ass because the sizing is based on width/height rather than bounding boxes.

I would normally chalk that up to growing pains of a new project, but reading through the GitHub issues and documentation didn't leave me with the impression that the author cared about how something could be done in Kitty, but only that it could be done in the most basic sense. If the user's workflow would benefit from having a partial overlay or popup, tough shit—they can either use a full overlay or create a layout for it.

It didn't sit well with me, and moving to Kitty full time would have been a downgrade in productivity for practically no real benefit.