We present a real-time procedure that uses cross-correlation and double-difference methods to rapidly relocate new seismic events with high precision relative to past events with accurately known locations. Waveforms of new events are automatically cross correlated with those archived for nearby past events to measure accurate differential phase arrival times. These data, together with delay times computed from arrival time picks, are subsequently inverted for the vector connecting the new event to its neighboring events using the double-difference algorithm. The new seismic monitoring technique is applied to earthquakes recorded in northern California, using near-real-time data feeds from the Northern California Seismic Network (NCSN) and the Northern California Earthquake Data Center, and a locally stored copy of the NCSN seismic archive. New events are automatically relocated in near-real time (tens of seconds) relative to a high-resolution double-difference earthquake catalog for northern California. Back testing using past events across northern California indicates that the real-time solutions are on average within 0.08 km laterally and 0.24 km vertically of the double-difference catalog locations. We show that the precision with which new events are located using this technique will improve with time, helped by the continued increase in density of recorded earthquakes and growth of the digital seismic archives. Real-time double-difference location allows for monitoring spatiotemporal changes in seismogenic properties of active faults with unprecedented resolution and therefore has considerable social and economic impact in the immediate evaluation and mitigation of seismic hazards.