Real-time crustal deformation monitoring is extremely important for achieving rapid understanding of actual earthquake scales, because the measured permanent displacement directly gives the true earthquake size (seismic moment, Mw) information, which in turn, provides tsunami forecasting. We have developed an algorithm to detect/estimate static ground displacements due to earthquake faulting from real-time kinematic GPS (RTK-GPS) time series. The new algorithm identifies permanent displacements by monitoring the difference of a short-term average (STA) to a long-term average (LTA) of the GPS time series. We assessed the noise property and precision of the RTK-GPS time series with various baseline length conditions and orbits and discerned that the real-time ephemerides based on the International GNSS Service (IGS) are sufficient for crustal deformation monitoring with long baselines up to ∼1,000 km. We applied the algorithm to data obtained in the 2011 off the Pacific coast of Tohoku earthquake (Mw 9.0) to test the possibility of coseismic displacement detections, and further, we inverted the obtained displacement fields for a fault model; the inversion estimated a fault model with M w 8.7, which is close to the actual Mw of 9.0, within five minutes from the origin time. Once the fault model is estimated, tsunami waveforms can be immediately synthesized using pre-computed tsunami Green's functions. The calculated waveforms showed good agreement with the actual tsunami observations both in arrival times and wave heights, suggesting that the RTK-GPS data by our algorithm can provide reliable rapid tsunami forecasting that can complement existing tsunami forecasting systems based on seismic observations.