Relation Extraction (RE) from Chinese historical documents plays a vital role in constructing historical knowledge bases and knowledge graphs, thereby safeguarding the invaluable cultural heritage preserved within these ancient texts. However, ancient texts written in classical Chinese pose significant challenges, including semantic complexity, limited annotated data, and domain-specific relation types that conventional “pre-training and fine-tuning” RE paradigms struggle to address effectively. To tackle these challenges, we propose HistoryRE, a reasoning-guided and knowledge-enhanced prompt learning approach. Our method designs a prompt template synthesis mechanism to guide the model in step-by-step reasoning, while leveraging external knowledge from classical literature to enhance the model’s comprehension of unique cultural and historical elements. Furthermore, we propose a regularization training mechanism based on entity masking to mitigate overfitting. Experimental results demonstrate that our approach achieves state-of-the-art relation extraction performance on two classical Chinese datasets, contributing to the digital preservation of China’s cultural legacy.