Brain-computer interface (BCI) systems based on motor imagery (MI) have been suggested as promising tools for the neurorehabilitation of stroke individuals. However, BCIs remain poor performers outside research environments and have been unable to reach the target public. This paper aimed to compile studies that evaluated motor imagery (MI)-based BCI intervention for stroke subjects, analyze the methodological quality of these studies, and verify the relationship between the effects of the interventions and performance achieved in MI tasks. Papers published between 2008 and 2020 were retrieved from five databases. The quality of the manuscripts was assessed using the Critical Review Form for Quantitative Studies. Fifteen articles met our eligibility criteria, with seven evaluated as excellent, four as very good, and four as good. Performance rates ranged from 58 to 90%, with two-thirds of studies achieving accuracies over 70%. Overall, studies reported a change in both the quality and the range of motion of the paretic limb after MI-BCI therapy, particularly when combined with conventional therapy. MI-BCI also assisted patients to assimilate tasks already forgotten by the brain as a result of stroke, and transfer this improvement to new circumstances. This review suggests that MI-BCI interventions may be a promising rehabilitation approach for stroke subjects.