A large portion of the cow's ration is composed of forages that can vary greatly in mineral concentrations, which may affect animal performance and health. Current methods for mineral analysis require sample destruction either through wet or dry ashing and complex analytical techniques for individual minerals. Energy dispersive X-ray fluorescence (EDXRF) is a nondestructive, multi-mineral, spectroscopy technique, which makes mineral quantification simple, direct, and affordable. The study objective was to evaluate the prediction performance of EDXRF of Na, Mg, P, S, Cl, Ca, K, Mn, Fe, Cu, and Zn concentrations in forages. Twelve certified plant samples with a wide range in mineral concentrations were used to develop calibrations, and 35 forages (18 alfalfa hays, 10 grass hays, 7 corn silages) with measured mineral concentrations, which were collected over several years from 3 proficiency programs, were used as an independent validation set. All the samples were previously dried and finely ground and were prepared by compressing them into a round dense pellet, 40 mm in diameter, using a 40-ton pneumatic laboratory press. Samples were scanned using an EDXRF instrument enhanced for lighter minerals like Na and Mg. Samples were scanned at 20 keV and at 40 KeV associated with an Al filter, for a total analysis time of approximately 6 min. Calibrations were developed with Bruker SpectraEDX (Bruker, Hamburg, Germany) software and optimized to minimize the standard error of calibration. All of the minerals had acceptable calibration performance with coefficient of determination ranging from 0.93 (P) and 0.99 (Cl, Ca, and Mn) and coefficients of variation within 5 to 14%, which are similar to the coefficients of variation of the reference analysis. The coefficients of variation for Na was an exception, with a coefficients of variation of 29%. The validation set obtained similar statistical results as that observed in calibration. The root mean square error of prediction corrected for bias was similar to the standard error of calibration, indicating that it is possible to build a robust calibrations that performed well across different type of forages by using 12 reference samples with a sufficient range in mineral concentrations that were determined accurately. A bias correction was necessary to improve prediction accuracy only for K (-0.23% dry matter) and Ca (-0.16% dry matter). Energy dispersive X-ray fluorescence demonstrated the ability to be an accurate, direct, and simple technique for forage mineral analysis.