Effectiveness of robotic gait training in rehabilitation of stroke patients remains inconclusive. A reason could be that the current robotic gait trainers do not initiate motor learning principles enough. To encourage active participation of the patient and therefore motor learning, assist-as-needed (AAN) support strategies have been implemented in the robotic gait trainer LOPESII. Aim of the current study was to examine the effect of assist-as-needed support on metabolic cost. Ten chronic stroke patients completed three 6-min walking trials in LOPESII, with zero support, AAN-support for stiff knee gait and complete-support. Metabolic parameters were measured and compared between support conditions. No significant differences in net metabolic power were observed between zero-support, AAN-support and complete support. No evidence was found that AAN-support asks a higher metabolic cost of the participant.