# Agent
MessagesState -> llm_call
llm_call -> should_continue(environment, END)
environment -> llm_call
