Abstract : The purpose of Argumentation is, in a knowledge base, for every proposition which validity has to be tested, to look for the reasons that support this proposition, as well as for the reasons that invalidate it. To this extent, a proposition is understood as a pair composed of both a proposition and the reason by which this proposition is meant to follow. Our purpose is to construct formal tools that allow the automatic generation of arguments by two agents having to face their respective knowledge. These tools rely on X-logics, a nonmonotonic formalism proposed by Siegel and Forget in 1996 as a general framework for achieving proofs in nonmonotonic reasoning. We show that the set X used to define the parameter of the inference relation gives the dynamic treatment of arguments an incomparable flexibility. After a survey of existing works related to logics and argumentation, X-logics are introduced, from which is composed the notion of attitude of an agent regarding a formula. Confrontation operators that map sets of formulas with the attitudes of the agent are then proposed. Further, a notion of answer that motivates the attitudes of the agent regarding a set of formulas is defined. Different forms of answers are distinguished, among which the notions of relevant answer, and lie. An answer representing the reasons that justify the conclusion of an argument, the computation of answers provides us with a method of automatic generation of arguments. Finally, we show that our argumentative framework can handle the generation of maximal conservative undercuts of Besnard and Hunter (2001), a kind of arguments especially interesting regarding relevance.