The well-documented eruptive history of Mount Mazama, Oregon, provides an excellent opportunity to use pre-eruptive volatile concentrations to study the growth of an explosive silicic magmatic system. Melt inclusions (MI) hosted in pyroxene and plagioclase crystals from eight dacitic–rhyodacitic eruptive deposits (71–7.7 ka) were analyzed to determine variations in volatile-element concentrations and changes in magma storage conditions leading up to and including the climactic eruption of Crater Lake caldera. Temperatures (Fe–Ti oxides) increased through the series of dacites, then decreased, and increased again through the rhyodacites (918–968 to ~950 to 845–895 °C). Oxygen fugacity began at nickel–nickel-oxide buffer (NNO) +0.8 (71 ka), dropped slightly to NNO +0.3, and then climbed to its highest value with the climactic eruption (7.7 ka) at NNO +1.1 log units. In parallel with oxidation state, maximum MI sulfur concentrations were high early in the eruptive sequence (~500 ppm), decreased (to ~200 ppm), and then increased again with the climactic eruption (~500 ppm). Maximum MI sulfur correlates with the Sr content (as a proxy for LREE, Ba, Rb, P2O5) of recharge magmas, represented by basaltic andesitic to andesitic enclaves and similar-aged lavas. These results suggest that oxidized Sr-rich recharge magmas dominated early and late in the development of the pre-climactic dacite–rhyodacite system. Dissolved H2O concentrations in MI do not, however, correlate with these changes in dominant recharge magma, instead recording vapor solubility relations in the developing shallow magma storage and conduit region. Dissolved H2O concentrations form two populations through time: the first at 3–4.6 wt% (with a few extreme values up to 6.1 wt%) and the second at ≤2.4 wt%. CO2 concentrations measured in a subset of these inclusions reach up to 240 ppm in early-erupted deposits (71 ka) and are below detection in climactic deposits (7.7 ka). Combined H2O and CO2 concentrations and solubility models indicate a dominant storage region at 4–7 km (up to 12 km), with drier inclusions that diffusively re-equilibrated and/or were trapped at shallower depths. Boron and Cl (except in the climactic deposit) largely remained in the melt, suggesting vapor–melt partition coefficients and gas fractions were low. Modeled Li, F, and S vapor–melt partition coefficients are higher than those of B and Cl. The decrease in maximum MI CO2 concentration following the earliest dacitic eruptions is interpreted to result from a broadening of the shallow storage region to greater than the diameter of subjacent feeders, so that greater proportions of reservoir magma were to the side of CO2-bearing vapor bubbles ascending vertically from the locus of recharge magma injection, thereby escaping recarbonation by streaming vapor bubbles. The Mazama melt inclusions provide a picture of a growing magma storage region, where chemical variations in melt and magma occur due to changes in the nature and supply rate of magma recharge, the timing of degassing, and the possible degree of equilibration with gases from below.