Since 2008, eruptive activity at the summit of Kı̄lauea Volcano, Hawai‘i has been confined to the new Overlook pit crater within the Halema‘uma‘u Crater. Among the broad range of magmatic processes observed in the new pit are recurring episodes of gas pistoning. The gas-piston activity is accompanied by seismic signals that are recorded by a broadband network deployed in the summit caldera. We use raw data recorded with this network to model the source mechanism of representative gas-piston events in a sequence that occurred on 20–25 August 2011 during a gentle inflation of the Kı̄lauea summit. To determine the source centroid location and source mechanism, we minimize the residual error between data and synthetics calculated by the finite difference method for a point source embedded in a homogeneous medium that takes topography into account. We apply a new waveform inversion method that accounts for the contributions from both translation and tilt in horizontal seismograms through the use of Green's functions representing the seismometer response to translation and tilt ground motions. This method enables a robust description of the source mechanism over the period range 1–10,000 s. Most of the seismic wavefield produced by gas-pistoning originates in a source region ∼1 km below the eastern perimeter of the Halema‘uma‘u pit crater. The observed waveforms are well explained by a simple volumetric source with geometry composed of two intersecting cracks featuring an east striking crack (dike) dipping 80°to the north, intersecting a north striking crack (another dike) dipping 65° to the east. Each gas-piston event is marked by a similar rapid inflation lasting a few minutes, trailed by a slower deflation ramp extending up to 15 min, attributed to the efficient coupling at the source centroid location of the pressure and momentum changes accompanying the growth and collapse of a layer of foam at the top of the lava column. Assuming a simple lumped parameter representation of the shallow magmatic system, the observed pressure and volume variations can be modeled with the following attributes : foam thickness (10–50 m), foam cell diameter (0.04–0.10 m), and gas-injection velocity (0.01–0.06 m s⁻¹). Gas-piston activity occurs in a narrow pipe with diameter of 6 m connecting the Halema‘uma‘u pit crater to the subjacent dike system. The height of the magma column is estimated at ∼104 m at the start of the sequence based on the period of very long period (VLP) oscillations accompanying the onset of the gas-piston signal. Based on the change in the period of VLP oscillations and tilt evidence, the height of the magma column is inferred to have risen by up to ∼23 m by the end of the 5 day long sequence. A penny-shaped crack model of the dike geometry yields effective diameters of ∼1.2–2.9 km for the east dike and 0.7 km for the north dike. The shallower north dike segment is embedded in a relatively weak medium, compatible with expected mechanical properties in the hydrothermal environment of this dike.