Most coding agent plans have a usage limit, and it has a way of running out in the middle of something rather than between tasks.
You have roughly three options. Wait until it resets. Pay for more. Or hand the rest to a different agent.
Waiting keeps everything the agent had worked out, but it might cost you hours. Switching gets you moving right away, but the second agent starts cold: it sees the repo and your prompt, and the first agent's partial work only if you pass it along.
In Styx, the app I'm building, I made the choice visible instead of burying it: when a plan is out, it shows when the limit resets, or sends the same ask to another agent.
So, when your plan runs out mid-task, what do you do? And if you switch agents, does the second one start from your original prompt, or from whatever the first one had done by then?
