Objective: The therapeutic alliance is possibly a crucial factor in treatment for borderline personality disorder (BPD). Among predictors of therapeutic alliance, aspects that have not yet been considered are metacognition or the patient's capacity to be aware of mental states. We therefore explored whether metacognition predicted alliance and if metacognition and therapeutic alliance together predicted outcome in brief treatment for BPD. Method: In a secondary analysis of a randomized controlled trial, we included N = 36 patients with BPD in the current study. The original trial assessed the effects of a 10 session psychiatric standard treatment with or without the added the Plan Analysis and the Motive Oriented Therapeutic Relationship. We assessed the therapeutic alliance session by session (Working Alliance Inventory), metacognition at session 1 (using the Metacognitive Assessment Scale-Revised) and outcome (using residual gains on the Outcome Questionnaire-45.2 between sessions 1 and 10). Results: A more differentiated capacity to understand the mind of the others at treatment onset predicted an increase of therapist-rated alliance over time. Therapist rated alliance was the only significant outcome predictor (B = -0.85, R Squared = .12). Conclusions: More differentiated metacognition predicted therapeutic alliance which in turn affected outcome, thus making metacognition a relevant therapy target early in therapy for BPD. Future studies should expand this investigation to patients with better functioning, treated with different modalities and with longer treatments.