Metacognitive Training for Depression (D-MCT), a low-Threshold group intervention, has been shown to improve depressive symptoms. It aims at the reduction of depression by changing dysfunctional cognitive as well as metacognitive beliefs. The purpose of the present study was to investigate whether the mechanisms of change in D-MCT are cognitive (and thus primarily concern the content of cognition) or metacognitive in nature. Eighty-four outpatients with depression were included in a randomized controlled trial comparing D-MCT to an active control intervention. Level of depression, dysfunctional cognitive beliefs (DAS), and metacognitive beliefs (MCQ subscales: Positive Beliefs, Negative Beliefs, Need for Control) were assessed before (t0) and after treatment (t1). Severity of depression was also assessed 6 months later (t2). Linear regression analyses were used to determine whether change in depression from t0 to t2 was mediated by change in cognitive vs. metacognitive beliefs from t0 to t1. D-MCT's effect on change in depression was mediated by a decrease in dysfunctional metacognitive beliefs, particularly 'need for control'. Our findings underline that one of the key mechanisms of improvement in D-MCT is the change in metacognitive beliefs. The current study provides further support for the importance of metacognition in the treatment of depression.