Post-stroke upper-limb motor recovery is often slow, substantially compromising patients' ability to perform activities of daily living and quality of life. As a novel closed-loop neuromodulation strategy, brain-computer interface (BCI) rehabilitation robots can decode neural activity signals in real time and convert them into control commands, enabling information interaction between the brain and the rehabilitation robot. Training with BCI rehabilitation robots can effectively improve upper-limb motor dysfunction in stroke patients. This article systematically reviews the key research progress, neural mechanisms, and existing challenges in the application of BCI rehabilitation robots for upper-limb motor recovery after stroke, aiming to provide references for researchers and clinicians in the fields of rehabilitation medicine, neural engineering, and robotics.