Abstract: To solve the problems of high manual participation, high time cost, low testing efficiency, and limited coverage in traditional User Interface (UI) testing, this paper proposed an intelligent UI testing method for railway passenger transport software based on large model and multi agent interaction. The paper designed a multi-agent interaction and collaboration mechanism for UI automatic testing, and analyzed the performance of current mainstream large models in testing intent understanding, test cases, and test script generation tasks based on practical cases. This method supports testers to propose testing requirements in natural language, automatically generate software testing scripts, and complete software testing tasks, provides effective reference for railway passenger transport software automation testing.