This paper provides an in-depth survey of LLM-based intelligent agents, focusing on their definitions, research frameworks, and foundational components. It covers both single-agent and multi-agent systems, detailing their planning, memory, rethinking, environment interaction, and action capabilities. The paper also explores the deployment of LLM-based agents in multi-agent systems, including multi-role collaboration, message passing, and strategies to address communication issues. Additionally, it discusses popular datasets and application scenarios, such as natural sciences, social sciences, engineering systems, and general domains. Finally, the paper concludes by envisioning the future prospects of LLM-based agents, considering the evolving landscape of AI and natural language processing.This paper provides an in-depth survey of LLM-based intelligent agents, focusing on their definitions, research frameworks, and foundational components. It covers both single-agent and multi-agent systems, detailing their planning, memory, rethinking, environment interaction, and action capabilities. The paper also explores the deployment of LLM-based agents in multi-agent systems, including multi-role collaboration, message passing, and strategies to address communication issues. Additionally, it discusses popular datasets and application scenarios, such as natural sciences, social sciences, engineering systems, and general domains. Finally, the paper concludes by envisioning the future prospects of LLM-based agents, considering the evolving landscape of AI and natural language processing.