Metallic phases and siderophile elements are critical to understanding the petrogenesis of the enigmatic ureilite meteorites. We obtained petrographic, major and minor element, and the first in situ trace element data for metallic phases (metal, sulfides, phosphide, carbide) in 24 main group ureilites of various petrographic types with Fo ~75-95. The most abundant type of metal (~1-3vol.%) occurs as ~10-40μm-wide strips along silicate grain boundaries. Ni contents of this metal range from ~0 to 7.3wt.% and are correlated with Co among all samples (Ni/Co=0.64×CI). A less abundant type of metal occurs as ~5-150μm diameter metallic spherules, consisting of cohenite (Fe3C), metal, phosphide and sulfide, enclosed in silicates (preferentially low-Ca pyroxene). Most samples contain 2 types of sulfide: (1) low-Cr (<0.1wt.%) troilite, and (2) lamellar intergrowths of daubreelite (FeCr2S4) and troilite.Abundances of 17 (mostly siderophile) elements were measured by LA-ICP-MS in grain boundary metal, spherules, graphite, sulfides and silicates. Average compositions of grain boundary metal in 10 samples show decreasing CI-normalized abundance with increasing volatility, interrupted by depletions in W, Mo, Ni and Zn, and enrichments in Au, As, Ga and Ge. CI-normalized Os abundances range from ~2 to 65, and are correlated with increasing Os/Pt, Os/Ni and Os/Pd ratios. CI-normalized Pt/Os ratios range from ~0.3 to 1. Bulk cohenite-bearing spherules have siderophile element abundances indistinguishable from those of grain boundary metal in the same sample. CI-normalized patterns of most siderophile elements in the metal are, within error, identical to those of the bulk rock (at 25-40× higher abundances) in each sample. There are no correlations between siderophile element abundances and Fo.We infer that at T≥1200°C ureilites contained immiscible Fe-C (3-4wt.% C) and Fe-S melts, small samples of which were trapped as the spherules within silicates. The Fe-S melt was largely extracted from the rocks, and the bulk of the residual Fe-C melt is now represented by the grain boundary metal. Assuming that ureilite precursor materials had CI or CV abundances of siderophile elements, the large fractionations of HSE observed in metal in 7 of the 10 samples require extremely high degrees (>98%) of batch Fe-S melt extraction, which implies very high xFeS (=wt. FeS/[Fe+FeS]) in the precursors. Furthermore, at such high degrees of fractionation, the HSE are so strongly concentrated into the residual metal that to match their relatively low absolute abundances in the ureilite metal, very high initial metal contents are required. Together, these constraints would imply that ureilite precursors had abundances of Fe metal and FeS (~20-35wt.% each) far exceeding those of known CC or OC. These requirements could be relaxed, permitting lower (more plausible) degrees of melting and lower initial metal and sulfide abundances, if ureilite precursors were volatile-depleted to a greater extent than bulk CV. We suggest that ureilite precursors contained, to various degrees, an overabundance (relative to chondrites) of refractory-enriched material such as CAIs. Excess CAIs could also account for observed depletions of W and Mo (otherwise difficult to explain) in the ureilite metal, and lead to the observed range of siderophile element patterns and abundances among samples. Such a model can potentially explain the lack of correlation between siderophile element abundances and FeO (or olivine Fo), and reconcile the metal and siderophile element data with a redox model for ureilite petrogenesis.