To provide more efficient usage of water resources and to minimize wastewater discharge, the design of water allocation networks has drawn increased attention over the past two decades. However, water system integration yields more structurally complex plants with greater interactions among the process units. Therefore, it is desirable to simplify the network structure with internal water mains, which help to attenuate the propagation of disturbances through the more complex water networks. In this paper, a general methodology for the design of optimal water-using networks with internal water mains is proposed using a new superstructure and a mixed-integer nonlinear programming (MINLP) strategy, given specifications of the mass load of the contaminants to be removed from each process unit and the lower and upper bounds on the contaminant concentrations. The mathematical program is formulated with constraints that bound the maximum number of outlet streams from each process unit. The most attractive networks, which have near-minimal freshwater consumption, relatively simple structures, and promising controllability, can be further designed to assess their operating and installation costs and controllability. This approach is applied to multicontaminant, water-using systems and is used for the design of water-using networks with multiple internal water mains.