MACRec is a novel multi-agent collaboration framework for recommendation systems, designed to enhance recommendation tasks through the collaborative efforts of specialized agents. Unlike previous work that focuses on simulating user or item behaviors, MACRec directly addresses recommendation tasks by deploying multiple agents with distinct roles and functions. These agents include the Manager, Reflector, User/Item Analyst, Searcher, and Task Interpreter, each contributing to the collaborative process. The framework provides customizable agents powered by large language models (LLMs) and useful tools, enabling developers to easily apply MACRec to various recommendation tasks such as rating prediction, sequential recommendation, conversational recommendation, and explanation generation. The framework and demonstration video are publicly available at https://github.com/wzf2000/MACRec. MACRec introduces a user-friendly online web interface that visualizes the agents' collaboration process, making it easier for developers to interact with the system. The framework's main strengths include a new multi-agent collaboration framework for recommendation, diverse applications on recommendation scenarios, and a user-friendly online web interface. The framework is designed to handle complex decision-making tasks in recommendation scenarios, where single-agent instances are unable to perform well. MACRec enables adaptable collaboration for various uses, supporting a wide range of recommendation scenarios through customizable configurations. The framework is the first open-source framework supporting multi-type agents for diverse recommendation scenarios.MACRec is a novel multi-agent collaboration framework for recommendation systems, designed to enhance recommendation tasks through the collaborative efforts of specialized agents. Unlike previous work that focuses on simulating user or item behaviors, MACRec directly addresses recommendation tasks by deploying multiple agents with distinct roles and functions. These agents include the Manager, Reflector, User/Item Analyst, Searcher, and Task Interpreter, each contributing to the collaborative process. The framework provides customizable agents powered by large language models (LLMs) and useful tools, enabling developers to easily apply MACRec to various recommendation tasks such as rating prediction, sequential recommendation, conversational recommendation, and explanation generation. The framework and demonstration video are publicly available at https://github.com/wzf2000/MACRec. MACRec introduces a user-friendly online web interface that visualizes the agents' collaboration process, making it easier for developers to interact with the system. The framework's main strengths include a new multi-agent collaboration framework for recommendation, diverse applications on recommendation scenarios, and a user-friendly online web interface. The framework is designed to handle complex decision-making tasks in recommendation scenarios, where single-agent instances are unable to perform well. MACRec enables adaptable collaboration for various uses, supporting a wide range of recommendation scenarios through customizable configurations. The framework is the first open-source framework supporting multi-type agents for diverse recommendation scenarios.