Building a multi-agent concierge system from scratch

In this video, Logan introduces a revamped multi-agent concierge system: a system that helps users achieve one of a set of tasks. The system consists of a set of agents that specialize in particular tasks, and an orchestrator agent that talks to the user and determines which agent is the best to help them achieve their task. Logan shows you how to build up the system step-by-step: * start with a basic agent * add tool use * implement human-in-the-loop to allow human approval of critical agent actions * expanding to multiple agents, mediated by an orchestrator It includes a shared global state between all the agents, plus a shared chat history.