The increased detection and treatment of early stage breast cancer as well as ductal carcinoma in situ (DCIS) has not led to significant survival benefits. Therefore, the current standard treatment of DCIS is questionable. An informed evidence-based treatment strategy, and likely de-escalation from the current standards requires new prognostic models built from more comprehensive characterization with objective criteria. Parallel profiling of the molecular landscape and micro-environment in pure DCIS remains challenging due to histological heterogeneity and the inevitable reliance on small archived specimens. Leveraging recent methodological advances, we characterized the mutational, transcriptional, histological and microenvironmental landscape across multiple micro-dissected regions from 39 cases to generate a multi-modal breast precancer atlas. The histological architecture was associated with grade, adiposity, and intrinsic expression subtypes. Similar to previous findings, high-grade lesions had higher mutational burden, including TP53 mutations, while low-grade lesions had more frequent 16q losses and GATA3 mutations. Multi-region analysis revealed most somatic alterations, including whole genome duplication events, were clonal, but genetic divergence increased with distance between regions. In 7/12 evaluable cases, somatic mutations in putative driver genes affected a subset of regions only. This genetic heterogeneity often accompanied phenotypic heterogeneity and regions with low risk features (Normal-like, Luminal A) occurred earlier than those with high-risk features (Her2-like, Basal or necrosis) according to the phylogenetic analysis. The immune-environment was evaluated using multiplex immuno-histochemistry to measure relative stromal and epithelial densities of B lymphocyte (B-cell), T lymphocyte (T-cell) and regulatory T cells (T-reg) and identify 3 immune-states: Active, Suppressed and Excluded (lower epithelial density). All states included both DCIS and adjacent benign regions, and none associated with intrinsic subtypes. The Excluded state was enriched in high-grade DCIS and, compared to benign areas, more likely acquired in DCIS, showing transcriptional evidence of stronger immune-suppression and possible evasion. The breast pre-cancer atlas therefore reveals correlated levels of phenotypic and genotypic heterogeneity, including at sub-histological resolution. These uniquely integrated observations will help scope future studies, prioritize candidate markers for progression risk modelling and identify functional similarities in precursor lesions from other types of adenocarcinomas.