Providing effective assistance with a powered exoskeleton requires fine synchronization of the artificial actuators with the residual and sound limb movement. In addition, it is fundamental to regulate the amount of energy generated by the assistive device at each step, as the energy required to walk changes with several factors, such as subject anthropometry, walking speed and cadence, and ground inclination. Conventional exoskeleton controllers do not allow direct regulation of energy generation and timing, but rather rely on experimental, subject-specific tuning. Consequently, outside of the specific tuning conditions—speed, cadence, and ground—the assistive energy level and timing are not optimal, thus the device cannot effectively improve walking. To overcome these limitations, we developed a novel energy-shaping controller based on Adaptive Frequency oscillators. The unique feature of this controller is that it can autonomously synchronize the assistive action of the robot with the user’s joints during periodic movements by relying on kinematics information only.