An in vivo approach combining high-throughput screening with cell type–specific readouts could enable elucidation of genotype-phenotype relationships in complex tissues. We developed an in vivo gain-of-function Perturb-seq platform, termed iGOF–Perturb-seq, to build a functional atlas of ~1000 transcription factors (TFs) in astrocytes, a cell type essential to many brain functions. We then identified cofunctional modules, annotated uncharacterized TFs, and predicted disease-associated TF clusters. Furthermore, iGOF–Perturb-seq performed in a mouse neuroinflammatory model identified Ferd3l as a therapeutic candidate, and astrocyte-specific overexpression of Ferd3l alleviated Alzheimer’s disease symptoms in mice. This study provides resources for understanding gene regulation and disease mechanisms in vivo and for identifying potential therapeutic targets for different brain diseases.