CRISPR-GPT is an innovative Large Language Model (LLM) agent designed to automate and enhance the design process of CRISPR-based gene-editing experiments. The system leverages the reasoning capabilities of LLMs, domain-specific knowledge, and external tools to facilitate the selection of CRISPR systems, design guide RNAs, recommend cellular delivery methods, draft protocols, and design validation experiments. CRISPR-GPT aims to bridge the gap between beginner biological researchers and advanced CRISPR genome engineering techniques by providing a user-friendly interface and detailed step-by-step guidance. The system consists of four core modules: LLM planner, Tool provider, Task executor, and LLM Agent. It supports a wide range of gene editing scenarios, including single gene knockout, base editing, prime editing, and epigenetic editing. CRISPR-GPT offers three interactive modes: "Meta Mode," which provides expert-defined pipelines for general gene editing scenarios; "Auto Mode," which automatically generates a tailored list of design tasks based on user input; and "Q&A Mode," which addresses ad hoc queries throughout the design process. Human evaluations by 12 experts in CRISPR and gene editing research showed that CRISPR-GPT outperformed general LLMs in terms of accuracy, reasoning, completeness, and conciseness. The system also demonstrated its efficacy through a real-world application, where it assisted in designing and performing a gene knockout experiment in human A375 cells. CRISPR-GPT addresses ethical and safety concerns by implementing mechanisms to prevent human heritable editing and protect user genome data privacy. The system ensures that users cannot bypass steps involving human tissues or organs and filters out identifiable long genome sequences to avoid leaking sensitive information. Overall, CRISPR-GPT showcases the potential of LLMs in automating complex biological experiments, providing researchers with a powerful tool to navigate the intricate landscape of CRISPR gene editing with ease and efficiency.CRISPR-GPT is an innovative Large Language Model (LLM) agent designed to automate and enhance the design process of CRISPR-based gene-editing experiments. The system leverages the reasoning capabilities of LLMs, domain-specific knowledge, and external tools to facilitate the selection of CRISPR systems, design guide RNAs, recommend cellular delivery methods, draft protocols, and design validation experiments. CRISPR-GPT aims to bridge the gap between beginner biological researchers and advanced CRISPR genome engineering techniques by providing a user-friendly interface and detailed step-by-step guidance. The system consists of four core modules: LLM planner, Tool provider, Task executor, and LLM Agent. It supports a wide range of gene editing scenarios, including single gene knockout, base editing, prime editing, and epigenetic editing. CRISPR-GPT offers three interactive modes: "Meta Mode," which provides expert-defined pipelines for general gene editing scenarios; "Auto Mode," which automatically generates a tailored list of design tasks based on user input; and "Q&A Mode," which addresses ad hoc queries throughout the design process. Human evaluations by 12 experts in CRISPR and gene editing research showed that CRISPR-GPT outperformed general LLMs in terms of accuracy, reasoning, completeness, and conciseness. The system also demonstrated its efficacy through a real-world application, where it assisted in designing and performing a gene knockout experiment in human A375 cells. CRISPR-GPT addresses ethical and safety concerns by implementing mechanisms to prevent human heritable editing and protect user genome data privacy. The system ensures that users cannot bypass steps involving human tissues or organs and filters out identifiable long genome sequences to avoid leaking sensitive information. Overall, CRISPR-GPT showcases the potential of LLMs in automating complex biological experiments, providing researchers with a powerful tool to navigate the intricate landscape of CRISPR gene editing with ease and efficiency.