This paper proposes a novel recommendation paradigm called Rec4Agentverse for LLM-based Agent platforms, aiming to provide personalized information services through the collaboration between Agent Items and Agent Recommender. The paradigm emphasizes the interaction and information exchange among users, Agent Recommender, and Agent Items, and is divided into three stages: User-Agent Interaction, Agent-Recommender Collaboration, and Agents Collaboration. In the first stage, the user interacts with Agent Items to obtain personalized information. In the second stage, Agent Items and Agent Recommender collaborate to provide personalized information services. In the third stage, Agent Items collaborate among themselves to further enhance personalized information services. The paper also discusses potential applications of Rec4Agentverse in various domains, such as travel, fashion, and sports, and explores research topics, including evaluation, preference modeling, and efficient inference. Additionally, the paper addresses potential issues and challenges, such as fairness and bias, privacy, harmfulness, robustness, and environmental friendliness. The paper concludes that Rec4Agentverse represents a new paradigm for recommendation systems that can significantly enhance the personalized information services provided by LLM-based Agent platforms.This paper proposes a novel recommendation paradigm called Rec4Agentverse for LLM-based Agent platforms, aiming to provide personalized information services through the collaboration between Agent Items and Agent Recommender. The paradigm emphasizes the interaction and information exchange among users, Agent Recommender, and Agent Items, and is divided into three stages: User-Agent Interaction, Agent-Recommender Collaboration, and Agents Collaboration. In the first stage, the user interacts with Agent Items to obtain personalized information. In the second stage, Agent Items and Agent Recommender collaborate to provide personalized information services. In the third stage, Agent Items collaborate among themselves to further enhance personalized information services. The paper also discusses potential applications of Rec4Agentverse in various domains, such as travel, fashion, and sports, and explores research topics, including evaluation, preference modeling, and efficient inference. Additionally, the paper addresses potential issues and challenges, such as fairness and bias, privacy, harmfulness, robustness, and environmental friendliness. The paper concludes that Rec4Agentverse represents a new paradigm for recommendation systems that can significantly enhance the personalized information services provided by LLM-based Agent platforms.