Objective The aim of the study was to demonstrate the feasibility, tolerability, and effectiveness of robotic-assisted arm training in incomplete chronic tetraplegia. Design Pretest/posttest/follow-up was conducted. Ten individuals with chronic cervical spinal cord injury were enrolled. Participants performed single degree-of-freedom exercise of upper limbs at an intensity of 3-hr per session for 3 times a week for 4 wks with MAHI Exo-II. Arm and hand function tests (Jebsen-Taylor Hand Function Test, Action Research Arm Test), strength of upper limb (upper limb motor score, grip, and pinch strength), and independence in daily living activities (Spinal Cord Independence Measure II) were performed at baseline, end of training, and 6 mos later. Results After 12 sessions of training, improvements in arm and hand functions were observed. Jebsen-Taylor Hand Function Test (0.14[0.04]–0.21[0.07] items/sec, P = 0.04), Action Research Arm Test (30.7[3.8]–34.3[4], P = 0.02), American Spinal Injury Association upper limb motor score (31.5[2.3]–34[2.3], P = 0.04) grip (9.7[3.8]–12[4.3] lb, P = 0.02), and pinch strength (4.5[1.1]–5.7[1.2] lb, P = 0.01) resulted in significant increases. Some gains were maintained at 6 mos. No change in Spinal Cord Independence Measure II scores and no adverse events were observed. Conclusions Results from this pilot study suggest that repetitive training of arm movements with MAHI Exo-II exoskeleton is safe and has potential to be an adjunct treatment modality in rehabilitation of persons with spinal cord injury with mild to moderate impaired arm functions.