Abstract: With the continuous advancement of digitalization and intelligence of railway passenger transport services in China, to improve the response capability of itinerary planning and recommendation systems in dynamic environments and enhance their adaptability to personalized user demands, this paper conducted research on the railway travel-oriented intelligent agent for itinerary planning and recommendation services. It designed and implemented an intelligent agent system for itinerary planning and recommendation services driven by the Large Language Model (LLM). By integrating Chain-of-Thought (CoT) prompting, Retrieval-Augmented Generation (RAG) and other technologies, the system realized the decomposition of complex user intentions and integrated reasoning of multi-source heterogeneous information. The paper adopted typical travel consultation scenarios to carry out simulation experiments, which preliminarily verified the system’s dynamic reasoning and personalized service adaptation capability in passenger itinerary planning and recommendation. The research provides a reference for constructing an efficient and personalized intelligent service system for railway travel.