In recent years, robots have taken place in surgical operations due to their advantages over humans, such as power, endurance, dexterity, and accuracy. Because of the lack of abilities, such as decision-making, adaptability, and creativity, human surgeons supervise the robots. The robots share the operation places with humans, called co-worker robots. Robots have the power to harm their environment; therefore, robots can generate dangerous situations for surgeons and patients. To deal with the issues, this dissertation aims to design active compliant control algorithms such as impedance control, admittance control, and hybrid position/force control to achieve safe interaction forces in surgical operations by considering the performance. The surgical co-worker robot’s type, actuation system, robot dynamics, and environment dynamics are important factors for designing the active compliant controller. Besides these, stability and robustness for safety, and agility and human effort for performance are considered for designing the controller. This dissertation takes into account three interaction scenarios encountered in surgical operations. In these scenarios, it is expected from the co-worker robot that it adapts to the sudden change in its environment dynamics. For instance, safe interaction is desired when the robot interacts with the stiff and soft tissues. To handle the issue, a switching control methodology is presented where the predefined control parameters are switched according to their environments. The methodology is implemented in a novel co-worker robot named NeuRoboScope, designed to assist the endoscopic pituitary gland surgery with the support of The Scientific and Technological Research Council of Turkey (TUBITAK). Moreover, active compliant control algorithms require a motion control algorithm as a low-level controller. In this dissertation, the computed torque method and independent joint controllers with gravity compensation are used as motion control algorithms. The computed torque method requires the dynamic model of the robot. Due to that, the dissertation proposes a simplified dynamic model with a correction coefficient for computational efficiency. ARM Cortex M4 processor runs the computed torque method with the proposed dynamic modeling method at 500 Hz. Also, this dissertation presents an independent joint controller which uses the simplified gravity matrix as a feedforward term for compensating the NeuRoboScope’s gravitational effect. The experimental results of both controllers are discussed in this dissertation.