Estimates of causal impacts can be needed to answer what-if questions about shifts in policy, such as new treatments in pharmacology or new pricing strategies for a business owner. In this paper we propose a non-parametric approach to estimate heterogeneous treatment effect based on random forests: HTERF. In the potential outcome framework with unconfoundedness we show that HTERF is pointwise a.s.-consistent to the true treatment effect. An interpretability result is also presented. A software implementation, CausalForest for Julia is available on the general repository of Julia.