The 2010 eruption of Merapi is the first large explosive eruption of the volcano that has been instrumentally observed. The main characteristics of the seismic activity during the pre-eruptive period and the crisis are presented and interpreted in this paper. The first seismic precursors were a series of four shallow swarms during the period between 12 and 4 months before the eruption. These swarms are interpreted as the result of perturbations of the hydrothermal system by increasing heat flow. Shorter-term and more continuous precursory seismic activity started about 6 weeks before the initial explosion on 26 October 2010. During this period, the rate of seismicity increased almost constantly yielding a cumulative seismic energy release for volcano-tectonic (VT) and multiphase events (MP) of 7.5 × 1010 J. This value is 3 times the maximum energy release preceding previous effusive eruptions of Merapi. The high level reached and the accelerated behavior of both the deformation of the summit and the seismic activity are distinct features of the 2010 eruption. The hypocenters of VT events in 2010 occur in two clusters at of 2.5 to 5 km and less than 1.5 km depths below the summit. An aseismic zone was detected at 1.5-2.5 km depth, consistent with studies of previous eruptions, and indicating that this is a robust feature of Merapi's subsurface structure. Our analysis suggests that the aseismic zone is a poorly consolidated layer of altered material within the volcano. Deep VT events occurred mainly before 17 October 2010; subsequent to that time shallow activity strongly increased. The deep seismic activity is interpreted as associated with the enlargement of a narrow conduit by an unusually large volume of rapidly ascending magma. The shallow seismicity is interpreted as recording the final magma ascent and the rupture of a summit-dome plug, which triggered the eruption on 26 October 2010. Hindsight forecasting of the occurrence time of the eruption is performed by applying the Material Failure Forecast Method (FFM) using cumulative Real-time Seismic Amplitude (RSAM) calculated both from raw records and on signals classified according to their dominant frequency. Stable estimates of eruption time with errors as small as ±4 h are obtained within a 6 day lapse time before the eruption. This approach could therefore be useful to support decision making in the case of future large explosive episodes at Merapi.