Diet and Fitness Recommender Using MCP and LLM

Abstract

The growing prevalence of chronic lifestyle diseases and increasing demand for personalized health management have created a pressing need for intelligent, adaptive fitness and dietary guidance systems. This paper presents a Diet and Fitness Recommender System developed using the Model Context Protocol (MCP) and Large Language Models (LLMs) to deliver personalized, context-aware health recommendations. The proposed system integrates a Flask-based web backend with an interactive frontend, enabling users to input personal parameters including age, weight, height, fitness goals, and dietary preferences. The system computes BMI, estimates daily caloric requirements, and generates dynamic, individualized meal and workout plans through LLM inference facilitated by MCP-structured context management. A gamified Just-In-Time Adaptive Intervention (JITAI) module for daily challenge tracking enhances user engagement. Evaluation demonstrates that the proposed system significantly outperforms traditional static recommendation approaches in personalization quality (4.3/5 satisfaction), nutritional accuracy (89%), and workout adherence (67%).