Most of the current graph contrast learning-driven recommender models tend to rely on a single view for training, which inevitably limits the ability of the models to fully capture the features of complex data. To this end, a recommendation algorithm multi-view knowledge contrastive learning recommendation (MKCLR) integrating multi-view contrastive learning and knowledge graph is proposed in this study. First, three view enhancement methods, namely, random edge discarding, adding uniform noise perturbation, and random walk algorithm, are used to construct three contrasting views for the knowledge graph and user-item graph. Second, LightGCN is used to encode the knowledge graph and construct multiple contrastive learning tasks, aiming to maximize the extraction and utilization of the rich information in the multi-view data. Finally, the main recommendation task is combined with contrastive learning for joint training. Experiments conducted on MIND, Last-FM, and Alibaba-iFashion show an average increase of 5.78% and 8.68% of MKCLR in terms of Recall and NDCG indexes, respectively, validating the effectiveness of the proposed method.