Although many studies have investigated mental workload during performance, its examination through functional connectivity during motor practice/learning is limited thus requiring further investigations. Therefore, this work aims to examine performance and functional connectivity dynamics underlying mental workload during motor practice by combining a robust computational method to derive connectivity and a human-machine interface which mitigates the use of participants’ prior motor experience since it can bias the acquisition process. Participants practiced reaching with a robotic arm through a head-controlled interface while kinematics and EEG were collected. The robotic end-effector kinematics quantified the performance and the Weighted Phase Lag Index indexed the connectivity during movement planning. Although performance improved during practice, the functional connectivity dynamics suggest that the recruitment of cognitive- motor resources decreased to a certain extent but that further training is likely needed to attenuate the mental workload. The work can also inform the training and design of assistive devices.