Background Characterized by high incidence rate, high disability rate, high mortality rate and high recurrence rate, stroke has become the second leading cause of death globally and the primary cause of adult disability. Though traditional rehabilitation methods have played a significant role in post-stroke functional recovery, their therapeutic efficacy is limited. In recent years, brain-computer interfaces (BCI) have advanced rapidly and are being used more frequently in rehabilitation training for stroke patients. Objective This systematic review and meta-analyses aimed to systematically assess the effect of brain-computer interface-based rehabilitation training on limb function in patients after stroke, and further investigate the efficacy differences among various types of brain-computer interfaces and treatment protocols. Methods The search strategy was conducted in 5 databases (PubMed, Scopus, Web of Science, Embase, and Cochrane Library databases) from inception to August 29, 2025. The studies that explored the impact of BCI combined with rehabilitation on limb function in stroke patients was mainly focused. A meta-analyses was performed using a random effects model, with the weighted mean difference (WMD) and 95% confidence intervals (CIs) as the effect sizes. Results Twenty-seven RCTs, including 23 that reported changes in upper limb function and four that reported changes in lower limb function, were included. The results showed that the training based on BCI significantly improved FMA-UE (Fugl-Meyer Assessment upper - extremity) [WMD = 3.50, 95% CI: (2.09, 4.90), p < 0.001] and FMA-LE (Fugl-Meyer Assessment lower -extremity) [WMD = 2.59, 95% CI: (1.94, 3.23), p < 0.001], compared with the control group. Conclusion The combined therapy was effective in improving the limb function of patients. BCI-based training might be a reliable rehabilitation program to improve limb function. Systematic review registration https://www.crd.york.ac.uk/PROSPERO/view/CRD420251038208 .