The pattern of eruptions at long-lived volcanic centers provides a window into the co-evolution of crustal magma transport, tectonic stresses, and unsteady magma generation at depth. Mount Mazama in the Oregon Cascades has seen variable activity over the last 400 ky, including the 50 km³ climactic eruption at ca. 7.7 ka that produced Crater Lake caldera. The physical mechanisms responsible for the assembly of silicic magma reservoirs that are the precursors to caldera-forming eruptions are poorly understood. Here we argue that the spatial and temporal distribution of geographically clustered volcanic vents near Mazama reflects the development of a centralized magma chamber that fed the climactic eruption. Time-averaged eruption rates at Mount Mazama imply an order of magnitude increase in deep magma influx prior to the caldera-forming event, suggesting that unsteady mantle melting triggered a chamber growth episode that culminated in caldera formation. We model magma chamber–dike interactions over ∼50 ky preceding the climactic eruption to fit the observed distribution of surface eruptive vents in space and time, as well as petrologically estimated deep influx rates. Best fitting models predict an expanding zone of dike capture caused by a growing, oblate spheroidal magma chamber with 10–30 MPa of overpressure. This growing zone of chamber influence causes closest approaching regional mafic vent locations as well as more compositionally evolved Mazama eruptions to migrate away from the climactic eruptive center, returning as observed to the center after the chamber drains during the caldera-forming eruption.