Terminal features you wish you had for AI development

Hey,

Since I started heavily using agents with CLI products, I've been super frustrated with Terminal. And looking at other options they mostly felt bloated and lacking features I wanted to make multi-agents piloting easier.

Of course, I embarked on building my own "Terminal" with basic things like tab naming, visual cues on tabs' content, branches, panes with text documents, clean notifications on end of runs, etc...

I will probably share that publicly and open source in a few days, but while I'm ironing things out I was wondering if while you were coding with Codex CLI or Claude there were features you wish you had.