[stingraycharles]

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.

[sothatsit]

In fact, proxy servers like LiteLLM[1] already exist that are very easy to setup and set limits in.

[1] LiteLLM: https://www.litellm.ai/