KRAS^G12C was recently identified to be potentially druggable by allele-specific covalent targeting of Cys-12 in vicinity to an inducible allosteric switch II pocket (S-IIP). Success of this approach requires active cycling of KRAS^G12C between its active-GTP and inactive-GDP conformations as accessibility of the S-IIP is restricted only to the GDP-bound state. This strategy proved feasible for inhibiting mutant KRAS in vitro; however, it is uncertain whether this approach would translate to in vivo. Here, we describe structure-based design and identification of ARS-1620, a covalent compound with high potency and selectivity for KRAS^G12C. ARS-1620 achieves rapid and sustained in vivo target occupancy to induce tumor regression. We use ARS-1620 to dissect oncogenic KRAS dependency and demonstrate that monolayer culture formats significantly underestimate KRAS dependency in vivo. This study provides in vivo evidence that mutant KRAS can be selectively targeted and reveals ARS-1620 as representing a new generation of KRAS^G12C-specific inhibitors with promising therapeutic potential.