Hand impairments following a stroke can impact their ability to perform daily activities including grasping, reaching and hand manipulation. Therefore, there is a need for more effective motor rehabilitation interventions post-stroke. Brain-Computer Interface (BCI) systems establish direct communication between the brain and the external environment, facilitating the repeated activation of closed-loop motor circuits damaged by stroke. Detecting a user’s hand movement intentions in EEG-based BCI systems remains one of the most challenging tasks within the BCI field. This study introduces a BCI-based dexterous hand rehabilitation robot for post-stroke grasping training, combining an EEG-vision hybrid intention recognition system with an 8-degree-of-freedom (DOF) hand rehabilitation glove.