LensKit provides a flexible framework for conducting offline evaluations of recommenders. Currently, train-test evaluation of recommender accuracy is supported; in the future, we will be adding additional evaluation capabilities.