Anyone else find that the hardest part of using local AI agents is actually invoking them mid-workflow?
I've got a few agents set up that I use regularly: one for summarizing, one I use while I'm in the middle of coding. They work well. The problem is getting to them without breaking whatever I'm already doing.
Alt-tabbing to a terminal or opening a new window every time kills the flow more than just doing the task manually would. I've tried keyboard shortcuts but mapping them cleanly across different apps without conflicts is its own project.
Started thinking about whether a dedicated physical trigger makes more sense than a hotkey for this kind of thing. Something that just fires the agent .
Curious if anyone else has thought about this or found a cleaner way to handle it.