Ground mass is redistributed during an earthquake causing the local gravitational potential to change. These gravitational fluctuations travel at the speed of light meaning they will arrive at a remote location significantly earlier than the fastest seismic waves. If these gravitational signals are measured by a gravimeter then early warning can be provided for an imminent earthquake. Earlier detection of earthquakes could be used to protect crucial infrastructure and save lives. The Torsion Pendulum Dual Oscillator (TorPeDO) is a gravity gradient sensor that has been constructed at the Australian National University. In this article we investigate the feasibility of measuring prompt gravitational transients from earthquakes with the TorPeDO. We simulated the response of the sensor to these signals and inserted these responses into scaled TorPeDO strain data to test their detection using a matched filter search. This simulation allows us to estimate the signal-to-noise ratio and detection time of the sensor to these transient signals, along with the influence of different detection thresholds on range and detection time. This article also proposes a method of earthquake localisation using TorPeDO sensors without the need for accurate signal timing. A real-time estimate of earthquake magnitude can be produced by combining this calculated location with TorPeDO strain data. We find that a TorPeDO system operating at design sensitivity would measure a moment magnitude 7.1 earthquake, 200~km away, reaching a signal-to-noise ratio of 5 at 15.7~s after the event starts. This will provide roughly 50.96~s of warning before the arrival of the first S waves.