Road infrastructure is extremely important in the rural areas of developing countries, where the main economic activity is agriculture. Using geo-referenced data on agricultural and transport sectors as well as instrumental variables specification based on the Kruskal algorithm we analyze the direct and indirect impact of road infrastructure on the agricultural sector in the Andes of Peru. The Andes mountains endow Peru with a very high topographical diversity. The high topographical diversity is related to longer travel time, higher road construction and maintenance costs, and lower density of the surface transportation network, which limits access to markets. We find that increasing road density in rural districts reduces the farmer's travel time by up to 2.9 hours as well the proportion of households that require less than 24 hours to travel from home to the district by up to 6 percent. Moreover, an additional unit of road density increases the participation in markets by 39 percent and reduces the proportion of output used for self-consumption and as own input by 49 and 5 percent respectively. Interestingly, the diversification of agricultural production follows an inverted U-shaped indicating that farmers reduce diversification as they experience the comparative advantage brought by the increased road connectivity. Our results are robust to the inclusion of year and region fixed-effects, and to controls for geography (i.e. gradient and altitude), population (i.e. education level), and mining activity.