Somatic mutations in the small GTPase K-Ras are the most common activating lesions found in human cancer, and such mutations are generally associated with poor response to standard therapies. Despite numerous efforts in academia and industry, small molecule inhibitors that directly target K-Ras remain elusive. Even more highly desired are molecules that selectively target mutant K-Ras while sparing the wild type protein. Efforts to directly target this oncogene have faced difficulties due to its picomolar affinity for GTP/GDP and the absence of known allosteric regulatory sites. Oncogenic mutations result in functional activation of Ras family proteins by impairing GTP hydrolysis. With diminished regulation by GTPase activity, the nucleotide state of Ras becomes more dependent upon relative nucleotide affinity and concentration. This gives GTP an advantage over GDP and increases the proportion of active GTP-bound Ras. We have used a fragment-based tethering screen followed by structure-guided chemical optimization to develop small molecules that irreversibly bind to the most frequent K-Ras mutant in lung cancer, K-Ras G12C. These compounds rely on the mutant cysteine for binding and therefore do not affect the wild type protein. Crystallographic studies reveal the formation of a new pocket that is not apparent in previous structures of Ras, beneath the effector binding switch-II region. Binding of these inhibitors to K-Ras G12C disrupts both switch-I and switch-II, subverting the native nucleotide preference to favor GDP over GTP and impairing binding to Raf. Our discovery of a new druggable pocket in K-Ras, and a set of inhibitors that bind to it in a mutant-specific fashion, provides a promising new avenue for the direct pharmacological inhibition of oncogenic Ras.