Large Language Models (LLMs) have tremendous promisesonconversational tasks in various sectors, including medical education. This study aims to integrate LLMs in medical education by user-centered iterative design and develop an educational clinical scenario simulator for clinical reasoning. The initial iteration prototypes a medical students-AI patient conversational app via prompt engineering. Feedback from physicians, student surveys, and focus group interviewsrevealed needs for a more comprehensive simulation mirroring the multi-agential nature of real clinical encounters. The second iteration prototypesan interactive LLM-based educational scenario simulator for clinical reasoning withan AI patient agent, multiple clinical dataacquisition agents, and educational assistant agents.Post-use surveysindicatetopfavouritesin clinical reasoningdevelopment(72.2%),real-time guidance(47.2%) and information gathering (44.4%). Theprogress from an LLM-powered conversational app to multi-agent educational simulator through iterative cycles with physicians and students-inputestablished a roadmapfor integrating LLMs into medical education and advancingAI-powered educational app design and development. Keywords