SecAct is a computational framework that infers the intercellular signaling activities of 1,170 human secreted proteins from spatial, single-cell and bulk transcriptomic data. It enables systematic analysis of secreted-protein-mediated intercellular communication across tissues and disease cohorts.