The breast tumor microenvironment is a complex ecosystem of epithelial, immune, endothelial, fibroblast and adipose cells. Cellular phenotypes and spatial localizations within the microenvironment are linked to patient prognosis and response to therapy. We established a protocol for multiplex imaging, cyclic immunofluorescence (CyCIF), and an image processing pipeline, mplexable, for quantitative single-cell spatial analysis of breast tumors and other tissues. We developed improved methods for signal removal, antibody specificity, background correction and batch normalization in CyCIF. We used our experimental and analytical methods for characterizing tumor heterogeneity and spatial relationships in diverse tissue samples including patient biopsies, surgical resections, tissue microarrays, cell lines and mouse models. We then generated CyCIF data and processed publicly available imaging mass cytometry and multiplex ion beam imaging data for biomarker discovery in large clinical cohorts of breast cancer patients. Our work reveals prognostic cell types and spatial arrangements in breast cancer as well as conserved tumor-stromal associations across breast cancer subtypes. Our CyCIF protocol, image analysis tools and atlases of breast cancer tissues all contribute to deepening our understanding of the tumor microenvironment in order to improve patient stratification and inform therapeutic strategies.