Abstract. We introduce, in this paper, a variational framework for the construction of improved treatment plans in volumetric modulated arc therapy for cancer treatment. This framework consists of a shape optimization component, from the molding of beam shapes, as well as strong constraints, from the equipment involved, on both beam shapes and intensities. We apply the binary level-set method, to handle complex shape topologies, and the fast sweeping method, to handle beam intensity constraints. The result is a fast, flow-based algorithm, in a simplified setting, that guarantees energy decrease. We include numerical simulations of clinical cases to demonstrate the efficacy of the approach.