I'm not sure it's even good though... the input doesn't need to come from a user. I have an "agent" which listens for an event in financial markets and then goe

I'm not sure it's even good though... the input doesn't need to come from a user. I have an "agent" which listens for an event in financial markets and then goes and does some stuff. In practice the current usage of "agent" is just: a program which does a task and uses an LLM somewhere to help make a decision as to what to do and maybe uses an LLM to help do it.