Microorganisms growing together within a community can either cooperate by exchanging nutrients or compete for the same nutrients. The resulting complex network of interactions leads to the emergence of various community functions. However, metabolic interactions are difficult to identify experimentally, and current computational predictions assume a community objective. Here, we propose a community objective-free computational method using a constraint-based metabolic model for predicting the minimal exchanges of nutrients between species coexisting in a chemostat at steady state. For a toy model, we showed that the species can be involved in different metabolic strategies, depending on the scarcity of resources and on their biomasses, and that alternate steady states exist. We then extended our method to larger-scale metabolic models and predicted the minimal interactions in a community of two amino acid auxotrophic E. coli strains. Overall, our approach proves promising for better identifying community interactions, for example, in the gut microbiome.