Kane Megamullion, an oceanic core complex near the Mid-Atlantic Ridge (MAR) abutting the Kane Transform, exposes nearly the full plutonic foundation of the MARK paleo-ridge segment. This provides the first opportunity for a detailed look at the patterns of mantle melting, melt transport and delivery at a slow-spreading ridge. The Kane lower crust and mantle section is heterogeneous, as a result of focused mantle melt flow to different points beneath the ridge segment in time and space, over an ∼300–400 kyr time scale. The association of residual mantle peridotite, dunite and troctolite with a large ∼1 km+ thick gabbro section at the Adam Dome Magmatic Center in the southern third of the complex probably represents the crust–mantle transition. This provides direct evidence for local melt accumulation in the shallow mantle near the base of the crust as a result of dilation accompanying corner flow beneath the ridge. Dunite and troctolite with high-Mg Cpx represent melt–rock reaction with the mantle, and suggest that this should be taken into account in modeling the evolution of mid-ocean ridge basalt (MORB). Despite early precipitation of high-Mg Cpx, wehrlites similar to those in many ophiolites were not found. Peridotite modes from the main core complex and transform wall define a depletion trend coincident with that for the SW Indian Ridge projecting toward East Pacific Rise mantle exposed at Hess Deep. The average Kane transform peridotite is a lherzolite with 5·2% Cpx, whereas that from the main core complex is a harzburgite with only 3·5% Cpx. As the area corresponds to a regional bathymetric low, and the crust is apparently thin, it is likely that most residual mantle along the MAR is significantly more depleted. Thus, harzburgitic and lherzolitic ophiolite subtypes cannot be simply interpreted as slow- and fast-spreading ridges respectively. The mantle peridotites are consistent with a transform edge effect caused by juxtaposition of old cold lithosphere against upwelling mantle at the ridge–transform intersection. This effect is far more local, confined to within 10 km of the transform slip zone, and far smaller than previously thought, corresponding to ∼8% as opposed to 12·5% melting of a pyrolitic mantle away from the transform. Excluding the transform, the overall degree of melting over 3 Myr indicated by the peridotites is uniform, ranging from ∼11·3 to 13·8%. Large variations in composition for a single dredge or ROV dive, however, reflect local melt transport through the shallow mantle. This produced variable extents of melt–rock reaction, dunite formation, and melt impregnation. At least three styles of late mantle metasomatism are present. Small amounts of plagioclase with elevated sodium and titanium and alumina-depletion in pyroxene relative to residual spinel peridotites represent impregnation by a MORB-like melt. Highly variable alumina depletion in pyroxene rims in spinel peridotite probably represents cryptic metasomatism by small volumes of late transient silica-rich melts meandering through the shallow mantle. Direct evidence for such melts is seen in orthopyroxenite veins. Finally, a late hydrous fluid may be required to explain anomalous pyroxene sodium enrichment in spinel peridotites. The discontinuous thin lower crust exposed at Kane Megamullion contrasts with the >700 km2 1·5 km+ thick Atlantis Bank gabbro massif at the SW Indian Ridge (SWIR), clearly showing more robust magmatism at the latter. However, the SWIR spreading rate is 54% of the MAR rate, the offset of the Atlantis II Fracture Zone is 46% greater and Na8 of the spatially associated basalts 16% greater—all of which predict precisely the opposite. At the same time, the average compositions of Kane and Atlantis II transform peridotites are nearly identical. This is best explained by a more fertile parent mantle beneath the SWIR and demonstrates that crustal thickness predicted by simply inverting MORB compositions can be significantly in error.