MXenes, derived from layered MAX phases, are a class of two-dimensional materials with emerging applications in energy storage, electronics, catalysis, and other fields due to their high surface areas, metallic conductivity, biocompatibility, and attractive optoelectronic properties. MXene properties are heavily influenced by their surface chemistry, but a detailed understanding of the surface functionalization is still lacking. Solid-state nuclear magnetic resonance (NMR) spectroscopy is sensitive to the interfacial chemistry, the phase purity including the presence of amorphous/nanocrystalline phases, and the electronic properties of the MXene and MAX phases. In this work, we systematically study the chemistry of Nb MAX and MXene phases, Nb2AlC, Nb4AlC3, Nb2CTx, and Nb4C3Tx, with their unique electronic and mechanical properties, using solid-state NMR spectroscopy to examine a variety of nuclei (1H, 13C, 19F, 27Al, and 93Nb) with a range of one- and two-dimensional correlation, wide-line, high-sensitivity, high-resolution, and/or relaxation-filtered experiments. Hydroxide and fluoride terminations are identified, found to be intimately mixed, and their chemical shifts are compared with other MXenes. This multinuclear NMR study demonstrates that diffraction alone is insufficient to characterize the phase composition of MAX and MXene samples as numerous amorphous or nanocrystalline phases are identified including NbC, AlO6 species, aluminum nitride or oxycarbide, AlF3·nH2O, Nb metal, and unreacted MAX phase. To the best of our knowledge, this is the first study to examine the transition-metal resonances directly in MXene samples, and the first 93Nb NMR of any MAX phase. The insights from this work are employed to enable the previously elusive assignment of the complex overlapping 47/49Ti NMR spectrum of Ti3AlC2. The results and methodology presented here provide fundamental insights on MAX and MXene phases and can be used to obtain a more complete picture of MAX and MXene chemistry, to prepare realistic structure models for computational screening, and to guide the analysis of property measurements.