Graph anomaly detection has important application value in network security, fraud detection and other fields, but the traditional supervised method limits its practical application due to the high cost of labeling and complex network structure. Although the unsupervised method can automatically identify anomalies, it is easy to be disturbed by noise and affect the detection accuracy. To solve these problems, we propose a Metapath-based Neighborhood Reconstruction Network (MNRN) for graph anomaly detection. MNRN reconstructs the relationship between nodes and their neighbors by designing specific meta-paths to capture context and connection patterns in complex graph structures. The model utilizes the prior information provided by a small amount of labeled data combined with a large amount of unlabeled data to effectively improve the accuracy and robustness of anomaly detection. The experimental results show that MNRN outperforms the current state-of-the-art methods on 6 real data sets, demonstrating its superior performance in anomaly detection tasks.