Most people with cerebral palsy (CP) suffer from impaired walking ability and pathological gait patterns. Seeking to improve the effectiveness of gait training in this patient population, this study developed and assessed the feasibility of a real-time biofeedback mechanism to augment untethered ankle exoskeleton-assisted walking performance in individuals with CP. We selected step length as a clinically-relevant gait performance target and utilized a visual interface with live performance scores. An adaptive ankle exoskeleton control algorithm provided assistance proportional to the real-time ankle moment. We assessed lower-extremity gait mechanics and muscle activity in seven ambulatory individuals with CP as they walked with adaptive ankle assistance alone and with ankle assistance plus step-length biofeedback. We achieved our technical validation goal by demonstrating a strong correlation between estimated step length and real step length (R = 0.771, p < 0.001). We achieved our clinical feasibility goal by demonstrating that biofeedback-plus-assistance resulted in a 14% increase in step length relative to baseline (p ≤ 0.05), while no difference in step length was observed for assistance alone. Additionally, we observed near immediate improvements in lower-extremity posture, moments, and positive power relative to baseline for biofeedback-plus-assistance (p < 0.05), with none, or more-limited improvements observed for assistance alone. Our findings suggest that providing real-time biofeedback and using step length as the target can be effective for increasing the rate at which individuals with CP improve their gait mechanics when walking with wearable ankle assistance.