This paper presents an algebra for conflict resolution in nonlinear planning. A set of conflicts in a plan is considered as a constraint network. Each node in the network represents a conflict, and is associated with a set of alternative ways for resolving it. Thus, resolving conflicts in a plan corresponds to selecting a set of consistent resolution methods so that, after they are applied to the plan, every conflict can be removed. The paper discusses the representional issues related to the conflict resolution, presents an algebra for resolving conflicts, and illustrates that some modified algorithms for preprocessing networks of constraints can greatly enhance the efficiency of conflict resolution.