Thoracoscopic left atrial appendage (LAA) occlusion is an alternative treatment for stroke prevention in patients with atrial fibrillation. Prospective study comparing thoracoscopic LAA occlusion and warfarin therapy is still lacking. The goal of this prospective cohort study was to assess the safety and efficacy of thoracoscopic LAA occlusion for stroke prevention in patients with nonvalvular atrial fibrillation compared with long-term warfarin therapy. Four hundred and ninety-two nonvalvular atrial fibrillation patients were enrolled. Two hundred and fifty-seven patients were treated with thoracoscopic LAA occlusion and 235 with long-term warfarin therapy. At 24 months, the rate of the first efficacy endpoint (composite of stroke, systemic embolism, and death) was 0.018 in the surgical group versus 0.043 in the warfarin group (p = 0.033). The rate of the second efficacy endpoint (stroke and systemic embolism excluding the first 7 days after procedure) was 0.010 versus 0.034 (p = 0.019). The rate of the first safety endpoint of bleeding was 0.016 versus 0.044 (p = 0.022). In conclusion, this study showed that thoracoscopic LAA occlusion was superior to warfarin for stroke prevention. The surgical group also had significantly lower bleeding risk. The incidence of surgical complications was low, and all occurred in hospital without causing serious outcomes.