CellAgent is an LLM-driven multi-agent framework designed for automated single-cell RNA sequencing (scRNA-seq) data analysis. It enables high-quality, automated processing of scRNA-seq data without human intervention. The framework includes three LLM-driven biological expert roles: Planner, Executor, and Evaluator. The Planner decomposes tasks, the Executor executes subtasks, and the Evaluator ensures solution quality through self-iterative optimization. CellAgent uses a hierarchical decision-making mechanism and a self-iterative optimization process to coordinate these roles, ensuring efficient and accurate task execution. CellAgent was evaluated on a comprehensive benchmark dataset comprising over 50 single-cell datasets from various tissues and cell types. Results showed that CellAgent consistently outperformed existing tools in tasks such as batch correction, cell type annotation, and trajectory inference. It achieved a 92% task completion rate, significantly higher than GPT-4 alone. CellAgent's automated framework reduces the technical and financial barriers for single-cell analysis, enabling more accurate and comprehensive biological discoveries. In batch correction, CellAgent demonstrated superior performance compared to other methods, achieving high scores in both batch correction and bio-conservation. For cell type annotation, CellAgent showed high accuracy, with 94% of clusters in the human PBMC dataset correctly annotated. In trajectory inference, CellAgent outperformed other methods, revealing developmental trajectories of human hematopoietic stem cells. CellAgent's architecture includes a memory module, tool retrieval, and code sandbox to manage tasks efficiently. It also employs self-evaluation mechanisms to optimize solutions iteratively. The framework is open-source and extensible, allowing users to provide new knowledge and tools to enhance its capabilities. CellAgent represents a significant advancement in automated single-cell data analysis, offering a robust solution for bioinformatics research and expanding the application of generative AI in scientific fields.CellAgent is an LLM-driven multi-agent framework designed for automated single-cell RNA sequencing (scRNA-seq) data analysis. It enables high-quality, automated processing of scRNA-seq data without human intervention. The framework includes three LLM-driven biological expert roles: Planner, Executor, and Evaluator. The Planner decomposes tasks, the Executor executes subtasks, and the Evaluator ensures solution quality through self-iterative optimization. CellAgent uses a hierarchical decision-making mechanism and a self-iterative optimization process to coordinate these roles, ensuring efficient and accurate task execution. CellAgent was evaluated on a comprehensive benchmark dataset comprising over 50 single-cell datasets from various tissues and cell types. Results showed that CellAgent consistently outperformed existing tools in tasks such as batch correction, cell type annotation, and trajectory inference. It achieved a 92% task completion rate, significantly higher than GPT-4 alone. CellAgent's automated framework reduces the technical and financial barriers for single-cell analysis, enabling more accurate and comprehensive biological discoveries. In batch correction, CellAgent demonstrated superior performance compared to other methods, achieving high scores in both batch correction and bio-conservation. For cell type annotation, CellAgent showed high accuracy, with 94% of clusters in the human PBMC dataset correctly annotated. In trajectory inference, CellAgent outperformed other methods, revealing developmental trajectories of human hematopoietic stem cells. CellAgent's architecture includes a memory module, tool retrieval, and code sandbox to manage tasks efficiently. It also employs self-evaluation mechanisms to optimize solutions iteratively. The framework is open-source and extensible, allowing users to provide new knowledge and tools to enhance its capabilities. CellAgent represents a significant advancement in automated single-cell data analysis, offering a robust solution for bioinformatics research and expanding the application of generative AI in scientific fields.