We provide a command, locmtest, that implements a test for exogeneity that is robust when the true relationship between the outcome variable and a discrete potentially endogenous variable is nonlinear. This test was developed in Lochner and Moretti (2015, Review of Economics and Statistics 97: 387–397), and it can be implemented even when only a single valid instrument is available. We present the motivation and general idea of the test. We also describe locmtest, which calculates the test, and provide empirical applications of the test and the command.