An automatic tool (named CLUSTER) for the prediction of the most probable crystal phases that can separate from glasses has been developed. The program analyzes the output of molecular dynamics simulations of glasses or glass ceramics, systematically sampling the ratios of the ions in different portions of the simulation box and comparing them to the stoichiometric ratio of compositionally equivalent crystalline phases retrieved from a crystal structure database. The efficacy of the similarity index elaborated has been judged by comparing the results obtained with the crystal phases identified by XRD analysis after thermal treatment in a series of multicomponent potential bioactive glasses and glass ceramics for which the advantages of rational-designed erosion-controlled release is straightforward.