This paper introduces generative agents, computational software agents that simulate believable human behavior. These agents can perform various daily tasks, form relationships, and coordinate activities, creating realistic social dynamics. The authors describe an architecture that extends a large language model to store and synthesize the agent's experiences, enabling dynamic planning and reflection. The architecture includes a memory stream for storing experiences, a retrieval function for selecting relevant memories, and a planning mechanism for generating actions. The agents are tested in a sandbox environment inspired by The Sims, where users can interact with a small town of 25 agents using natural language. Evaluations show that the agents produce believable individual and emergent social behaviors, such as spreading invitations for a Valentine's Day party and coordinating group activities. The paper also discusses the ethical and societal risks of generative agents and suggests ways to mitigate these risks.This paper introduces generative agents, computational software agents that simulate believable human behavior. These agents can perform various daily tasks, form relationships, and coordinate activities, creating realistic social dynamics. The authors describe an architecture that extends a large language model to store and synthesize the agent's experiences, enabling dynamic planning and reflection. The architecture includes a memory stream for storing experiences, a retrieval function for selecting relevant memories, and a planning mechanism for generating actions. The agents are tested in a sandbox environment inspired by The Sims, where users can interact with a small town of 25 agents using natural language. Evaluations show that the agents produce believable individual and emergent social behaviors, such as spreading invitations for a Valentine's Day party and coordinating group activities. The paper also discusses the ethical and societal risks of generative agents and suggests ways to mitigate these risks.