A planning problem mainly consists of a description of an initial world, a set of formulas which declaratively define a goal world and a set of operator specifications which can be used to derive new worlds from given worlds. The goal of the planning procedure is to find a sequence of operators such that if they are applied to the initial world in succession, they result in a goal world in which all the formulas of the goal description hold. Deductive databases can be described by a triple consisting of a set of facts, a set of ru_les, and a set of constraints that each database state has to obey. Deductive databases will be used to model worlds in a planning domain. Further, we assume the existence of a consistency maintenance procedure which allows to find for an inconsistent database state consistent ones. The exploitation of automatic consistency maintenance is the main issue of the pa­ per. Hence, it will be described first. A simple example will motívate the usefulness of consistency maintenance for planning. Then, a planning procedure and sorne variants are specified which build upon consistency maintenance. This planner mainly consists of two layers, the deductive database and a specialized procedure devoted to planning. We call this a hybrid or heterogeneous approach. The basic planning procedure as well as sorne more advanced ones will be specified and sorne theoretical and empirical results will be g1ven. While this planner outperforms existing ones in both, expressiveness and performance, there still exist sorne problems which have to be tackled in order to improve its perfor­ mance. This will lead to a new, homogeneous planner. The key idea is to axiomatize the planning procedure and use consistency maintenance to derive descriptions of possible plans.