A BCI-controlled hand exoskeleton activates neuroplasticity mechanisms, promoting motor learning. The contribution of perception to this phenomenon is understudied. The aim of this study was to assess the impact of sensorimotor integration on the effectiveness of neurorehabilitation based on the learning of a hand opening movement by stroke patients using BCI and to investigate the effect of ideomotor training on spasticity in the paretic hand. The study was conducted in 58 patients (median age: 63 (22; 83) years) with traumatic brain injury, ischemic (76%) or hemorrhagic (24%) stroke in the preceding 2 (1.0; 12.0) months. The patients received 15 (12; 21) ideomotor training sessions with a BMI-controlled hand exoskeleton. Hand function was assessed before and after rehabilitation on the Fugl–Meyer, ARAT, Frenchay, FIM, Rivermead, and Ashworth scales. An increase in muscle strength was observed in 40% of patients during flexion and extension of the radiocarpal joint and in 29% of patients during the abduction and adduction of the joint. Muscle strength simultaneously increased during the abduction and adduction of the radiocarpal joint (p < 0.004). Ideomotor training is ineffective for reducing spasticity because no statistically significant reduction in muscle tone was detected. Improved motor performance of the paretic hand was positively correlated with improvements in daily activities. Motor training of the paretic hand with a robotic orthosis activates kinesthetic receptors, restores sensation and improves fine motor skills through better sensorimotor integration.