This paper presents algorithms for automated design of multishot molds for manufacturing multimaterial or multicolor objects, and focuses on molding planning that determines a sequence of mold stages required to produce the desired object. By modeling a multimaterial object as an assembly of homogenous components, a geometric reasoning approach is proposed to generate feasible or practical mold stage sequences by combining the assembly planning approach and the two-plate mold design method. First, a graph-based assembly model, namely the attributed contact graph, is derived from the B-rep models of the constituent components of the gross object by detecting and representing all the contacts between mating components explicitly. Then, all feasible mold stage sequences, represented by an AND/OR graph, are generated by reasoning on geometric constraints due to the demoldability and connectedness requirements using an assembly-by-disassembly strategy. Depending on its demoldability, each component is to be made by one of the three basic molding strategies with varied mold stages and/or mold pieces. To narrow the choice, an optimal or practical molding plan is searched from the feasible molding plans according to some criteria such as the number of mold stages, the number of side cores, and flatness of the parting line. Finally, starting from the last mold stage, mold pieces for each mold stage of the selected molding plan are constructed recursively. The feasibility of the proposed algorithms is demonstrated through an implemented prototypical system, which has been tested successfully with various multi-material objects.