The identification of patients with metastatic triple-negative breast cancer (mTNBC) who are expected to benefit from platinum-based chemotherapy is of interest. We conducted a single-arm phase II clinical trial of single-agent platinum for mTNBC with biomarker correlates. Patients with mTNBC received first- or second-line cisplatin (75 mg/m(2)) or carboplatin (area under the concentration-time curve 6) by physician's choice once every 3 weeks. Coprimary end points were objective response rate (RR) and response prediction by p63/p73 gene expression. Secondary and exploratory end points included toxicity assessment, RR in cisplatin versus carboplatin, and RR in molecularly defined subgroups, including BRCA1/2 mutation carriers. Patients (N = 86; 69 as first-line therapy) received cisplatin (n = 43) or carboplatin (n = 43). RR was 25.6% (95% CI, 16.8% to 36%) and was numerically higher with cisplatin (32.6%) than with carboplatin (18.7%). RR was 54.5% in patients with germline BRCA1/2 mutations (n = 11). In patients without BRCA1/2 mutations (n = 66), exploratory analyses showed that a BRCA-like genomic instability signature (n = 32) discriminated responding and nonresponding tumors (mean homologous recombination deficiency-loss of heterozygosity/homologous recombination deficiency-large-scale state transitions HRD-LOH/HRD-LST scores were 12.68 and 5.11, respectively), whereas predefined analysis by p63/p73 expression status (n = 61), p53 and PIK3CA mutation status (n = 53), or PAM50 gene expression subtype (n = 55) did not. Five of the six long-term responders alive at a median of 4.5 years lacked germline BRCA1/2 mutations, and two of them had increased tumor HRD-LOH/HRD-LST scores. Platinum agents are active in mTNBC, especially in patients with germline BRCA1/2 mutations. A measure of tumor DNA repair function may identify patients without mutations who could benefit from platinum therapy agents. Prospective controlled confirmatory trials are warranted.