This survey provides an overview of the progress and challenges in Large Language Model (LLM)-based Multi-Agent systems. LLMs have demonstrated strong reasoning and planning capabilities, making them suitable for autonomous agents. Recent advancements have led to LLM-based multi-agent systems that can solve complex problems and simulate real-world environments. These systems leverage the collective intelligence and specialized skills of multiple agents, enabling collaborative planning, discussion, and decision-making. The survey discusses the essential aspects of LLM-based multi-agent systems, including the agents-environment interface, agent profiling, communication, and capability acquisition. It also summarizes the commonly used datasets and benchmarks for these systems and maintains an open-source GitHub repository to track the latest research. The survey highlights the applications of LLM-based multi-agent systems in various domains, such as software development, embodied agents, science experiments, and world simulations. It also addresses the challenges in this field, including the need for multi-modal environments, addressing hallucination, acquiring collective intelligence, and scaling up LLM-based multi-agent systems. The survey concludes with a discussion on the future research directions and opportunities in this rapidly evolving field.This survey provides an overview of the progress and challenges in Large Language Model (LLM)-based Multi-Agent systems. LLMs have demonstrated strong reasoning and planning capabilities, making them suitable for autonomous agents. Recent advancements have led to LLM-based multi-agent systems that can solve complex problems and simulate real-world environments. These systems leverage the collective intelligence and specialized skills of multiple agents, enabling collaborative planning, discussion, and decision-making. The survey discusses the essential aspects of LLM-based multi-agent systems, including the agents-environment interface, agent profiling, communication, and capability acquisition. It also summarizes the commonly used datasets and benchmarks for these systems and maintains an open-source GitHub repository to track the latest research. The survey highlights the applications of LLM-based multi-agent systems in various domains, such as software development, embodied agents, science experiments, and world simulations. It also addresses the challenges in this field, including the need for multi-modal environments, addressing hallucination, acquiring collective intelligence, and scaling up LLM-based multi-agent systems. The survey concludes with a discussion on the future research directions and opportunities in this rapidly evolving field.