This paper proposes a novel framework for robotic manipulation tasks, exploiting the Human-Robot Collaboration (HRC) potential. The framework integrates two adaptive controllers to i) modulate robot compliance in contact with the environment along constrained directions, and to ii) enable human guidance through touch when a manual intervention is needed. To demonstrate the potential of the proposed framework, we consider a collaborative screwing task. In this example application, the operator is in charge of placing the screws on the table and following the instructions on a graphical user interface. The robot, after identifying the position of the screws through an online human pose-tracking system, performs the screwing using the proposed controller. The human operator can adjust the screwing position of the robot using the adaptive interface at anytime if the position accuracy through vision is insufficient. We first experimentally evaluate the operation of the proposed controller and demonstrate its performance in comparison to the classical impedance control. Next, the overall system is evaluated in a collaborative (human and robot) setting.