Background: Trastuzumab (T), a monoclonal antibody (mAb) targeting HER2, is standard of care palliative 1^st-line therapy for advanced HER2+ GEJ/GC patients (pts). M, an Fc-engineered anti-HER2 mAb, targets the same HER2 epitope but with higher affinity for both 158V (high binding) and 158F (low binding) alleles of activating Fc receptor CD16A. M coordinately enhanced both innate and adaptive immunity, including antigen-specific T-cell responses to HER2. PD-1 and LAG-3 are T-cell checkpoint molecules that suppress T-cell function. MGA012 (INCMGA00012) is a humanized, hinge-stabilized, IgG4 κ anti-PD-1 mAb blocking binding of PD-L1 or PD-L2 to PD-1. MGD013 is a humanized Fc-bearing bispecific tetravalent protein that binds to both PD-1 and LAG-3, inhibiting their respective ligand binding. We previously reported that a CTX-free regimen of M+PD-1 blockade was well tolerated in GEJ/GC pts, and induced a 30% objective response rate (ORR). This was 2- to 3-fold greater than in historical controls with checkpoint inhibitors alone. This registration-directed trial assesses efficacy, safety, and tolerability of M+checkpoint inhibition ± CTX in metastatic/locally advanced, treatment-naïve, HER2+ GEJ/GC pts. Methods: This is a 2-cohort, adaptive open-label phase 2/3 study. The first single arm, CTX-free cohort A evaluates M+MGA012 in HER2+ (immunohistochemistry [IHC] 3+) and PD-L1+ (excluding microsatellite instability high) pts. After 40 pts are evaluated for response/safety, 60 more pts will be enrolled if the threshold for continuation is met. In randomized cohort B, HER2+ (IHC 3+ or 2+/fluorescent in situ hybridization+) pts are enrolled irrespective of PD-L1 status. Part 1 randomizes pts to 1 of 4 arms (50 pts each): control arm (T+CTX) or 1 experimental arm (M+CTX; M+CTX+MGA012; M+CTX+MGD013). CTX is investigator’s choice XELOX or mFOLFOX-6. Part 2 consists of control (T+CTX) vs 1 experimental arm (M+CTX) + either MGA012 or MGD013, depending on results from part 1; with 250 pts each. The primary efficacy endpoint for cohort A (both parts) is ORR per RECIST 1.1; for cohort B part 2 it is overall survival.