Monolithic simulation approaches, while often providing high accuracy, may face some limitations. These include differences in time-scales between subsystems, varied solver requirements, high costs of system redesign, and limited access to internal subsystem data. As an alternative approach, co-simulation enables different simulation tools to run simultaneously for interconnected physical systems. This is particularly relevant when the simulation includes a mechanical multibody system that involves unilateral contact interactions. Our primary interest here is the multi-domain systems where at least one subsystem is a mechanical multibody system with contact interactions. In a co-simulation framework, a system is divided into subsystems that exchange information at predetermined \emph{communication points}. Each subsystem is connected to other subsystems through \emph{interface}. Real-time simulations for virtual environments and prototyping would need non-iterative co-simulation, for which stability is often the main problem. The stability behaviour is controlled by how the system elements are coupled. Model-based coupling offers a potential solution for enhancing simulation accuracy and efficiency. It involves using a reduced model of a mechanical subsystem, with the interfacing subsystem, whether mechanical or from another domain, interacting only with this model. A key element is that the reduced model must accurately represent the mechanical subsystem’s behavior. In this context, the reduced model is developed to represent some specific behaviour of a larger scale subsystem. A reduced interface model is developed by transforming the generalized velocities of the mechanical system to the subset of interface velocities. This would decompose the motion space into the interface subspace and its orthogonal complement that is factored out using mass-orthogonal decoupling. Such a reduced interface model (RIM) characterizes the specific behaviour that is associated with the interface [1]. The main challenge in developing the reduced models often emerges when the system is non-smooth, i.e., subjected to friction/unilateral interactions. In this work, we discuss the development of a representative reduced order model for a non-smooth mechanical subsystem in a multi-domain setup. To this end, a hydraulically actuated crane model is selected as the case study, where the system is divided into two subsystems: the hydraulic actuation system and the crane multibody.