In this article, a dynamic Principal–Agent model with discrete actions is analysed from a Multi-Objective optimization framework. As a result, a concave Pareto Frontier is numerically approximated. The concavity of the Pareto Frontier is a consequence of the information asymmetry between the Principal and the Agent. The underlying Multi-Objective framework allows us to consider more powerful assumptions than those used in the traditional Single-Objective optimization approach. As contracts move in the Pareto Frontier (trade-off surface) towards those that are more advantageous to the Agent, the prevalence of compensation plans in which the Principal assumes most of the risk of the productive activity are observed. When the Principal and the Agent are more patient, both obtain higher values of their discounted expected utilities, which generates a higher level of economic surplus. The Agent faces lower variability in future compensation when it is costlier for him to exert an additional effort unit. Finally, a new Multi-Objective Evolutionary Algorithm (MOEA) is proposed in this article to approximate Pareto Frontiers, such algorithm involves an innovative ranking-mutation mechanism which promotes approximations with good spread, achieving even better results that some obtained by already well known MOEAs.