Wouldn't the obvious solution to this problem to stop using agents that don't respect your usage limits instead of trying to build sketchy containers around misbehaving software?
Yeah I don’t understand this problem, who uses so many agents at the same time in the first place?
And if this is really a problem, why not funnel your AI agents through a proxy server which they all support instead of this hacky approach? It would be super easy to build a proxy server that keeps track of costs per day/session and just returns errors once you hit a limit.
And if this is really a problem, why not funnel your AI agents through a proxy server which they all support instead of this hacky approach? It would be super easy to build a proxy server that keeps track of costs per day/session and just returns errors once you hit a limit.