In Spring 1999, the first experiments in real-time oceanic forecasting were performed during at-sea operations on Georges Bank. The scientific mission was focused on cross-frontal exchange processes in tidal mixing fronts as part of the U.S. GLOBEC program. Computational teams representing six institutions (Dartmouth College, Woods Hole Oceanographic Institution, National Marine Fisheries Service, the University of North Carolina, Brookhaven National Laboratories and the Bedford Institute of Oceanography) were deployed simultaneously on 2 ships and a shore station. 3-day forecasts of circulation and Lagrangian transport on the Bank were performed on a daily schedule, assimilating basin-scale weather and oceanic pressure forecasts plus local drifter, ADCP, CTD and VPR observations. The time lag achieved between data availability and forecast was 12 hours for the atmospheric forecast, and 3-4 hours for local observations. Shipboard services included forecasts of Eulerian circulation, mixing, Lagrangian trajectories, and animated frontal dynamics, These products were focused on the problem of locating frontal features and tracking dye releases and observed plankton distributions. Skill was measured as the precision in forecasting trajectories of drifters associated with dye release experiments. Error growth in the ensemble of 'best' forecasts averaged 3 Km/day, in a circulation regime with tidal displacements of 5-10 Km and mean subtidal speeds of 20-30 cm/sec. VPR observations of 2 planktonic species were assimilated and their forecast trajectories illustrate persistent separation of shoalward predators and seaward prey.