Access to quality and comprehensive healthcare is a prevalent issue in resource-poor and linguistically diverse areas. To tackle this, we introduce MediCURE, a disease prediction AI system that combines interpretable machine learning, Retrieval Augmented Generation (RAG), and multi-language capability to provide contextualized and personalized health advice. MediCURE’s innovation comes in the form of its hybrid structure: a transparent, symptom-based disease prediction is done using a Decision Tree Classifier, whereas a RAG module directly extracts appropriate home remedies and medical knowledge from a hand curated knowledge base. The system is accessed through a conversational chatbot that dynamically gathers user input, iteratively refines predictions based on follow-up queries, and produces elaborate visual and textual explanations in the user’s target language. MediCURE was tested with a symptom-disease structured dataset and compared against baseline classifiers, where the Decision Tree resulted in 91.97% accuracy. MediCURE’s potential as a highly scalable, real-time AI assistant that closes the gap between predictive analytics and accessible healthcare, especially for underserved populations, is demonstrated in the results.