Single-cell studies have revealed substantial microglial diversity in development, homeostasis and disease. However, a framework enabling comparison and stratification of microglial states across contexts is needed. Here we generated an atlas of myeloid cell states by single-cell RNA sequencing more than one million central nervous system cells from more than 30 physiological and pathological conditions. This atlas enables us to establish a comprehensive taxonomy of myeloid cell states across brain disorders and related mouse models, comprising 27 superclusters and 192 clusters that are prevalent across diseases and largely conserved. We augment this taxonomic framework with spatial transcriptomics to map how immune cell states are organized within tissue and interact with their local cellular environment. Using in vivo perturbations, we also show that activation-associated microglial states are dependent on interferon and colony-stimulating factor 1 receptor signaling. Together, these findings provide a spatially aware taxonomic framework for central nervous system immune cells in health and disease.