Oncolytic virus therapy has shown activity against primary melanomas; however, its efficacy in brain metastases remains challenging, mainly because of the delivery and immunosuppressive nature of tumors in the brain. To address this challenge, we first established PTEN-deficient melanoma brain metastasis mouse models and characterized them to be more immunosuppressive compared with primary melanoma, mimicking the clinical settings. Next, we developed an allogeneic twin stem cell (TSC) system composed of two tumor-targeting stem cell (SC) populations. One SC was loaded with oncolytic herpes simplex virus (oHSV), and the other SC was CRISPR-Cas9 gene-edited to knock out nectin 1 (N1) receptor (N1
KO
) to acquire resistance to oHSV and release immunomodulators, such as granulocyte-macrophage colony-stimulating factor (GM-CSF). Using mouse models of brain metastatic BRAF
V600E
/PTEN
−/−
and BRAF
V600E/wt
/PTEN
−/−
mutant melanomas, we show that locoregional delivery of TSCs releasing oHSV and GM-CSF (TSC-G) activated dendritic cell– and T cell–mediated immune responses. In addition, our strategy exhibited greater therapeutic efficacy when compared with the existing oncolytic viral therapeutic approaches. Moreover, the TSCs composed of SC-oHSV and SC
N1KO
–releasing GM-CSF and single-chain variable fragment anti–PD-1 (TSC-G/P) had therapeutic efficacy in both syngeneic and patient-derived humanized mouse models of leptomeningeal metastasis. Our findings provide a promising allogeneic SC-based immunotherapeutic strategy against melanomas in the CNS and a road map toward clinical translation.