Neoproterozoic kimberlite, ultramafic lamprophyre, and carbonatite magmatic activity was widespread across the Canadian-Greenland Shield. Models to explain the preponderance of this deeply-derived CO2-rich magmatism between 680–540Ma range from impingement of multiple mantle plumes to rifting activity linked to the breakout of the Laurentian plate from the Rodinia supercontinent configuration. We add to the debate about the origin of kimberlite magmas and evaluate possible mantle sources of the 655Ma ‘diamond-rich’ Renard (new SIMS U/Pb perovskite ages) and 629Ma ‘barren’ Wemindji kimberlites on the eastern Superior craton in Quebec, Canada.Our Sr-Nd-Hf and carbon isotope data (87Sr/86Sri=0.70241–0.70442; εNdi=+0.2 to +4.8; εHfi=+0.3 to +6.5; δ13C=−5.6 to −3.9‰) suggest a common and moderately depleted convecting upper mantle source region for both the Renard and Wemindji kimberlites, which occur 400 km apart in the interior of the Superior craton. In contrast, the low Os isotope ratios (187Os/188Osi=0.11078–0.12620; γOsi=−13.7 to −1.6) and unfractionated chondritic relative HSE abundances (Os, Ir, Ru, Pt, Pd, Re) indicate significant involvement of ancient refractory cratonic mantle material in kimberlite magma formation. Our model calculations suggest that for both the diamond-rich Renard and the barren Wemindji kimberlite magmas up to 30% of the Os was derived from refractory cratonic peridotites. This material might have been assimilated by originally more CO2-rich carbonated silicate melts derived from the asthenosphere. We also show that the geochemical and Sr-Nd-Hf-Os isotopic compositions of the Renard and Wemindji kimberlites do not require significant input from melts derived from olivine-poor cratonic mantle lithologies such as MARID-type veins and pyroxenites/eclogites. This contrasts with the petrogenesis of deeply-derived volatile-rich potassic magmas found along the peripheries of cratons (e.g., ultramafic lamprophyres, kamafugites, and olivine lamproites), a setting where abundant non-peridotitic components have been added to the lithospheric mantle over the course of continent evolution.Provided that CO2-rich melts, such as proto-kimberlites, occur near the solidus of volatile-fluxed peridotites, no excess mantle heat is required in their formation. This important but often overlooked constraint, together with the observation that there exist no spatial or temporal relationships between the Superior craton kimberlites and Large Igneous Provinces during the Late Neoproterozoic, suggests that kimberlite magmatic activity was tectonically controlled. In our preferred model, ubiquitous CO2-rich proto-kimberlite melts form during volatile-controlled redox melting processes at ambient mantle temperatures in a thermal boundary layer directly beneath thick cratonic lithosphere. The success rate of ‘evolving’ hybrid kimberlite magmas reaching Earth’s surface increases when tensile stresses propagate into the >200km thick keels of continental lithosphere. These conditions are frequently met during fast and changing plate motions associated with the assembly and breakup of supercontinents.