This paper presents a comprehensive survey on the memory mechanisms of large language model (LLM)-based agents. The authors argue that memory is a critical component that differentiates LLM-based agents from traditional LLMs, enabling them to perform complex, long-term interactions with environments. The survey discusses the importance of memory in LLM-based agents from three perspectives: cognitive psychology, self-evolution, and agent applications. It also reviews existing memory mechanisms, their design, and evaluation methods, and presents various applications of memory-enhanced agents. The authors highlight the limitations of current memory mechanisms and propose future research directions, including more advanced parametric memory, memory in multi-agent applications, lifelong learning, and memory in humanoid agents. The survey provides a systematic overview of memory mechanisms in LLM-based agents, offering insights for future research and development.This paper presents a comprehensive survey on the memory mechanisms of large language model (LLM)-based agents. The authors argue that memory is a critical component that differentiates LLM-based agents from traditional LLMs, enabling them to perform complex, long-term interactions with environments. The survey discusses the importance of memory in LLM-based agents from three perspectives: cognitive psychology, self-evolution, and agent applications. It also reviews existing memory mechanisms, their design, and evaluation methods, and presents various applications of memory-enhanced agents. The authors highlight the limitations of current memory mechanisms and propose future research directions, including more advanced parametric memory, memory in multi-agent applications, lifelong learning, and memory in humanoid agents. The survey provides a systematic overview of memory mechanisms in LLM-based agents, offering insights for future research and development.