We have developed an experimental system for minimally invasive surgery providing force feedback and automation of recurring task. The system consists of four robotic arms, which can be equipped with either minimally invasive instruments or a stereo camera.The master console provides a stereo view of the field of operation and two input devices can feed back forces to the user. We have utilized this system to assess the possibility of automating difficult handling tasks like surgical knot tying. In order to achieve this, a novel approach for human-machine skill transfer was developed. It constitutes an extension to learning by demonstration, which is a well known paradigm of robotic learning.