Background Non-epithelial ovarian tumours encompass a heterogeneous group of neoplasms comprising germ cell tumours (GCT) and sex-cord stromal tumours (SCST). These tumours are characterized by an extensive inter- and intratumoral heterogeneity. To better understand the landscape of immune cells in non-epithelial ovarian cancer, we construct a blueprint of stromal heterogeneity using single-cell RNA sequencing (scRNA-seq). Methods We performed scRNA-seq of 66 919 cells from fresh tissue collected from 12 patients, both in primary and recurrent setting. Most tissue samples were derived from SCST (n=9), including 7 adult granulosa cell tumours, 1 primary juvenile granulosa cell tumour and 1 primary Sertoli-Leydig cell (SL) tumour. Three samples were obtained from treatment-naïve GCT (2 immature teratomas (IT) and one dysgerminoma (DG)). We characterise immune cell subtypes phenotypically by highlighting its specific marker genes. Results Based on differential expression analysis and expression of transcriptomic markers, we identified 27 clusters consisting of 9 tumour cell and 18 stromal cell clusters. By clustering the transcriptome of 5961 T-cells and natural killer (NK) cells, T and NK cells deriving from DG (58%) and Sertoli-Leydig cell (20%) were found to be predominantly expressed. Remarkably, the DG tumour sample exhibited a pronounced cytotoxic environment, dominated by CD8+ exhausted T-cells whereas in the SL tumour a more naïve condition was observed. Half of the B-cells, mainly in differentiated cell states, derived from the DG sample. A nearly absence of B cells in all granulosa cell tumors. Regarding the myeloid lineage, the granulosa cell samples were dominated by LYVE1 and CX3CR1 macrophages. The latter are known to be inversely correlated with T cell expansion and thus predictors of low response to anti-PD-1 treatment. Conclusions With this analysis, we generate a comprehensive blueprint of the tumour micro-environment in non-epithelial ovarian cancer. Despite a limited sample size, we obtain high-resolution insights. The cytotoxic environment of dysgerminoma and immune desert profile of granulosa cell tumours are novel findings, which, however, need to be validated in larger datasets.