Personalized nutrition is an approach to generate food recommendations based on individuals gut microbiome profile and several other factors. Personalized nutrition recommendation system based on individuals gut microbiome, knowledge graphs, graph neural networks (GNNs), and reinforcement learning (RL) is presented in this paper. Developing a knowledge graph by integrating data on gut microbiome, nutrients, metabolites, and disease associations serves as the base for the personalized recommendation system. The nodes and edges relationship defined in the KG acts as base layer for the GNN algorithms to capture information to generate recommendations and RL algorithms updates recommendations based on feedback loop. The system demonstrated improved results in three-phase validation tests, and first phase demonstrated an 87% accuracy while predicting the dietary recommendations and 83% accuracy in microbiome-diet associations. The second phase test used synthetic clinical profile which exhibited 91% accuracy. The third test phase was an expert evaluation that showed an 89% accuracy.