Primary brain tumors and peripheral cancers with central nervous system (CNS) metastases are often fatal despite best available therapies. Recognizing PD-L1 as a shared target across high-risk primary and metastatic CNS tumors, we explored whether our PD-L1:CD3 T-cell engager (TCE) could effectively eliminate brain cancers or metastases despite the presence of PD-L1 on some non-neoplastic cells (e.g., pancreatic islet cells, lung epithelial cells, leukocytes). We evaluated anti-tumor activity against 14 intracranial xenograft models and observed extended survival in 13, including models with both high and low initial PD-L1 expression. Across individual xenografts monitored by bioluminescence, 92% responded to TCE treatment. In a model of treatment-refractory lung intracranial metastases, TCE was efficacious at 0.03-1.5 mg/kg and had superior activity to osimertinib and atezolizumab. In PD-L1 humanized mice, TCE was tolerated at all tested doses with only grade 1-2 toxicities observed. Dose range finding in non-human primates revealed pharmacodynamic responses at all doses and grade 2-4 toxicities observed at the highest 0.3mg/kg dose. TCE administration did not adversely affect function or morphology of PD-L1-positive lung or pancreatic cells and effect on leukocytes was limited. This PD-L1:CD3 TCE presents a therapeutic candidate for patients with fatal tumors.