Endometrial cancer (ENDOM) prevention remains challenging, creating an urgent need for better risk stratification tools. We developed a prompt-based bimodal multilevel endometrial cancer (2M-EC) predictive platform that integrates clinically accessible data with multi-biofluid molecular profiles through MALDI-TOF MS analysis. Our study established the MBF-ED cohort (n=531), collecting comprehensive clinical data and multi-dimensional body fluid samples. We processed these using a unique analytical pipeline: (1) simplifying clinical variables through empirical and data-driven methods, (2) extracting ENDOM-specific MS features using machine learning, and (3) developing an innovative bimodal AI architecture that fuses 2D MS omics matrices with 1D clinical vectors. The resulting prompt-based 2M-EC predictive platform provides real-time, interpretable risk stratification through an online interface. The advantages of it include overcoming single-marker limitations via multimodal integration, combining molecular depth with clinical practicality and scalable design adaptable to both resource-limited and advanced healthcare settings. This work demonstrates how AI can bridge cutting-edge molecular profiling with routine clinical practice, offering a new paradigm for task-oriented cancer risk assessment.