The National Weather Service's Meteorological Development Laboratory (MDL) has developed ensemble-based probabilistic MOS guidance. The product, known as Ensemble-Kernel Density MOS (EKDMOS), incorporates output from the North American Ensemble Forecast System (NAEFS). NAEFS is a multi-model ensemble comprised of the Global Ensemble Forecast System (GEFS) and the Canadian Meteorological Centre Ensemble (CMCE). EKDMOS provides statistically post-processed forecasts of 2-meter temperature, 2-meter dew point, daytime maximum temperature, and nighttime minimum temperature. The forecasts include a mean forecast and several points on the cumulative distribution function (CDF) which describes the forecast uncertainty. We present a method that develops station-based spread-skill functions that relate ensemble-member spread to the expected ensemble-mean error variance. These so-called spread-skill functions are used to objectively calibrate the final EKDMOS forecast CDF to match the degree of forecast uncertainty. EKDMOS is currently awaiting operational implementation. Once the implementation is complete, gridded EKDMOS forecasts covering the CONUS and Alaska will be publicly disseminated via the NWS's National Digital Guidance Database (NDGD). We also present a comparative, station-based verification of EKDMOS, the operational GFS MOS, and bias-corrected downscaled forecasts from the NAEFS. EKDMOS reduced the 2-meter temperature mean absolute error (MAE) by 2-11% compared to the GFS MOS and by 1-19% compared to the downscaled NAEFS product, depending on projection. Averaged over all projections EKDMOS reduced the MAE by 6% and 9% compared the GFS MOS and downscaled NAEFS product respectively. Comparisons of the continuous ranked probability score (CRPS) and probability integral transform (PIT) histograms indicate that EKDMOS produces well calibrated probabilistic forecasts that are more statistically reliable than existing products.