KREEP is a lunar material having very high concentrations of incompatible elements; its name is an acronym for the incompatibles K, rare-earth elements (REE), and P. Although a few pristine (endogenously igneous) KREEPy samples were returned from the Apollo 15 and 17 sites, most KREEPy samples are polymict breccias. Most models of KREEP petrogenesis have been based on partial melting of a variety of sources. Such models fail to explain the veritable absence of variations in incompatible element patterns over the sampled portion of the moon. We have defined a KREEP component based on the average composition of Apollo 14 breccias having extremely high concentrations of incompatible elements. Normalization of accurate incompatible data for KREEPy samples from the Apollo 12, 14, 15, 16, and 17 sites to this component virtually always shows no resolvable fractionation (e.g., <10% variation in the La/Lu ratio), whereas partial melting models typically produce larger fractionations (±20–25% in La/Lu) from a factor of 2 difference in degree of partial melting. Required is a single major source that could provide KREEP to widely separated locations on the nearside of the moon. The anorthositic crust of the moon is commonly attributed to the flotation of plagioclase on a deep, moon-wide magma ocean. Fractional crystallization of this magma ocean would have produced large enrichments of incompatibles in a residual liquid. No other plausible major source of incompatibles has been proposed. We borrow the German prefix ur—meaning primeval and designate this residual liquid ‘urKREEP.’ We propose that all KREEPy rocks originated by dilution of urKREEP with crustal or mantle materials during assimilation, or zone-refining (pristine samples), or impact-induced brecciation (breccias and melt rocks). The formation of urKREEP cannot be dated precisely. Correction of breccia Rb-Sr model ages for Rb loss or gain during the early intense bombardments yields ages that cluster in the range 4.4–4.5 Gy. This implies that crystallization of the magma ocean was essentially complete at this time and is in general agreement with U-Pb evidence indicating crustal formation at 4.4 Gy. Assuming that the moon had the composition of an H-group chondrite depleted in Fe-Ni and FeS and that half the incompatibles fractionated into materials other than urKREEP, the thickness of a moon-wide urKREEP layer was <2 km. Thorium concentrations determined by gamma ray spectroscopy indicate that about 4% of the incompatibles in an H chondritic moon are now in the outermost kilometer.