Purpose/Significance To build an agent for medical record generation with large language model (LLM) as the core, so as to alleviate the burden of clinical medical record writing. Method/Process Data for model fine-tuning are prepared through data preprocessing and large language model simulated dialogues generating. A medical record generation rule library is established based on medical record writing standards. A specialized medical record generation model is trained through model fine-tuning, combining technologies such as knowledge retrieval, prompt engineering, and workflow design, and then a medical record generation agent is built. Result/Conclusion The electronic medical records(EMR) generated by the agent have a higher accuracy rate and are in line with the norms of EMR and the requirements of medical quality control, providing empirical references and support for the application and development of intelligent medical record generation.