Abstract: A novel approach to learning first order logic formula with constraints from positive and negative examples and background knowledge is presented. Adding with our new method to learn constraints to the ILP system Progol, the new system can derive a CLP program, covering all positive examples and consistent with all negative examples, by comparison between the positive and negative examples, without the user's hint and intervention. This paper presents this new CILP system and some experiments, and points out some directions of the future research.