We study the adoption of an environmental technology -- demi-lunes -- in Niger. Like many environmental technologies, demi-lunes require an upfront investment in exchange for medium-run benefits, which agronomists estimate to be substantial. We implement a cluster randomized control trial in 180 villages with treatments designed to relax informational, credit and labor constraints. Relative to a pure control, training increases the probability of adoption by 90 percentage points. Combining training with either unconditional or conditional cash transfers has no additional effect on the extensive margin of adoption, but increases the intensity of adoption by 35-50 percent relative to training alone. We also observe increases in agricultural output, consistent with agronomic descriptions of the costs and benefits of adoption, as well as other measures of household well-being. Over 90 percent of treatment households have operational demi-lunes two years later. Using the pattern of results and our experimental design, we investigate the mechanisms underlying our findings.