@Michael Guiao Hi! Thinking about it from a 'how do we execute actions' point of view, I'm building out a foundation of agents that very specifically can handle all functions for a single tool or api as a baseline. With the idea agents that can act as project managers higher up the hierarchy that can call on those agents with a specific order of operations necessary to execute a task. Also, several 'project managers' can share the specific tool agents. As an example: 1 agent can execute all gmail functions 1 agent can execute all google calendar functions 1 agent can execute all zoom functions .... then I have another gmail agent that just reads incoming messages and decides what to do with them. So, gmail reader passes an email to a scheduling agent. then the 'scheduling agent' can call on the 3 example agents above. - email the availability - select the time - schedule the meeting - create and add the zoom link The idea is that if you have agents that can execute all the tools you use to their fullest, then capabilities should be able to scale as AI agent reasoning/logic increases. If you build out a workflow for specific functions, you'll always be building workflows

@Nate Herk I expanded on your email manager to add a check for spam but to verify with me first over telegram. I was thinking of dropping these into pinecone so I can later re-train the model and remove the human feedback. But I thought it might be valuable for a bit. wondering what you thought