Faced with a vast and highly unstructured body of historical archives, traditional digitization and keyword retrieval methods are insufficient to meet the increasingly complex needs of historical research.This is especially true when dealing with complex historical materials like The Sheng Xuanhuai Archives,which involve a multi-dimensional network of late Qing Dynasty politics, economy, and diplomacy.Researchers often find themselves in a predicament of incomplete retrieval, unclear organization, and difficulty in making connections.This study, based on a review of the technological background of the digital humanities' transformation from digitization and structuring to intelligentization, introduces cutting-edge thinking chain technology and reasoning-action framework from large language models to construct an "Intelligent Analysis System for The Sheng Xuanhuai Archives" based on Agentic RAG.This system breaks through the limitations of traditional predefined knowledge graphs, achieving semantic vectorization and dynamic reasoning of the entire archive text through the collaborative work of five agents: "intent understanding, data retrieval, data summarization, evaluation and decision-making, and content writing." The system possesses the capabilities of proactively planning retrieval paths, multi-step logical deduction, cross-document evidence verification, and self-correction, simulating the cognitive process of historians: "proposing hypotheses—collecting historical materials—verifying and analyzing—forming conclusions." This article demonstrates the system's application potential through typical cases at the micro, meso, and macro levels.Research practice proves that this "human-in-the-loop" collaborative model in the AI era not only liberates historians from tedious information searching but also opens up new possibilities for historical research characterized by digital archaeology and holographic logic enhancement.