As an alternative to the frequently used mixed boundary conditions in ocean GCM's, we present a dynamic atmospheric model (ECBILT) that is simple and yet describes the relevant dynamic and thermodynamic feedback processes to the ocean. This provides the possibility of studying atmosphere/ocean dynamics on very long time-scales of the order of a thousand years. The model is two orders of magnitude faster than AGCMs. We have been running ECBILT with prescribed SSTs for a period of 500 years with seasonal cycle included both in the solar forcing and in the climatological SSTs. The mean state and the variability properties of ECBILT are reasonably realistic. The simulation of the surface fluxes is quite realistic and justifies coupling ECBILT to an ocean GCM. We have done two SST anomaly experiments, one with a tropical SST anomaly as observed during January 1983 and one with an SST anomaly in the northern extra-tropical Atlantic ocean. For the tropical SST anomaly experiment the amount of anomalous precipitation agrees well with what has been found with atmospheric GCM's, but the resulting extra-tropical teleconnection pattern is very weak. The atmospheric response pattern to extra-tropical SST anomalies agrees well with similar SST anomaly experiments performed with atmospheric GCM's. We have tested the performance of ECBILT in coupled mode by coupling it to a simple ocean GCM and thermodynamic sea-ice model and integrating the coupled system for a period of thousand years after a spin up of 6000 years. The simulation of the mean water mass distribution and the mean circulation of the ocean resembles the observed ocean circulation. It has a warm and fresh bias and the circulation and associated transports are too diffuse and too weak. The ocean's variability is realistic, considering the simplicity of the ocean model, although a bit too weak. We have explored the covariability between the atmosphere and ocean over the Northern Atlantic ocean by performing a singular value decomposition of SST anomalies and 800 hPa geopotential height anomalies. The 2nd mode shows a peak in both spectra at a timescale of about 18 years. The time scale of this mode is set by the ocean but the physical mechanisms that are operating are not yet unambiguously identified. The simulation of this coupled extra tropical decadal mode of variability, which also shows up in the observations and in much more complex coupled models provides strong evidence for the potential usefullness of ECBILT when studying atmosphere/ocean interaction and the associated ocean variability on time scales ranging from decades to many thousands of years.