This paper presents intelligent object distribution architecture to maximize the performance and intelligence of a distributed surgical robotics system and its preliminary implementation in an MR-guided surgical robot system in an open-configuration MR! scanner. The method enables networked integration of a robot control server and multiple clients with minimum engineering overhead but maximum flexibility and performance. The clients in this study include an intraoperative imager, high-performance image processing computers), and surgical navigation host. The first contribution of the paper is to propose the use of object distribution by common object request broker architecture (CORBA), in which a robot control object on the robot control server can be remotely but transparently invoked from the clients regardless of their hardware, operating systems, or programming language. Second, we propose a technique to achieve additional flexibility by reporting the robot configuration information, i.e. geometry and kinematics of the robot, to the clients upon connection. Third, we ensure protection against an unauthorized entity by introducing a security control host that authorized the clients’ access to the robot server. In a prototype implementation of an MR-guided surgical robot system, the robot was controlled by surgical navigation software (the 3D Slicer) on a UNIX client by invoking the distributed control object on a robot control server on a PC. The method was evaluated in performance studies; and the result indicated 3.6 milliseconds for retrieving positions of the robot stages and 25.5 milliseconds to send a frame-based motion command, which are satisfactory for surgical robot control. In conclusion, the proposed method shows the potential usefulness of flexibly integrating the legacy software to a surgical robot system with minimum engineering overhead, thereby achieving highly complex and intelligent tasks in robot-assisted surgery.