Multimodal recommendation systems leverage multimodal information such as text and visuals to introduce additional knowledge, and typically achieve better recommendation accuracy compared to traditional knowledge graph-based recommendation systems. Multimodal information is often integrated into item representations or used to uncover underlying structures, capturing hidden relationships between different modalities. This paper proposes a multimodal knowledge graph recommendation method based on Multi-view Cross Contrastive Learning (MVCC). MVCC constructs multiple views based on multimodal knowledge graph information, user-item interaction information, and kno-wledge graph information. It utilizes graph neural networks to learn user and item representations under different views and fuses these views to obtain the final user and item representations. Additionally, MVCC employs contrastive learning to perform cross-view alignment, enabling the different views to align with each other, thereby capturing more comprehensive and accurate features of items and users, and improving the generalization ability of the model. We conduct evaluation experiments on two real-world datasets, and the results show that MVCC significantly outperforms other classic recommendation models, demonstrating its effectiveness.