The parameter setting of an algorithm that will result in optimal performance is a tedious task for users who spend a lot of time fine-tuning algorithms for their specific problem domains. This paper presents a multi-agent tuning system as a framework to set the parameters of a given algorithm which solves a specific problem. Besides, such a configuration is generated taking into account the current problem instance to be solved. We empirically evaluate our multi-agent tuning system using the configuration of a genetic algorithm applied to the root identification problem. The experimental results show the validity of the proposed model.