Insurance fraud in vehicle insurance is a substantial issue that results in inflated costs, financial losses for insurers and honest policyholders, and potential economic instability. This paper presents a MERN (MongoDB, Express.js, React, Node.js) web application designed to automate the claims process and enhance the detection of fraudulent insurance claims. The system enables customers to file claims digitally, uploading necessary documents and providing relevant details, these details are then processed by a trained machine learning model to classify claims as fraudulent or valid. The dataset used consists of 15,419 instances with 33 features underwent data preprocessing techniques, such as ADASYN and SMOTE, to address class imbalance. Various machine learning models, including Random Forest, K-Nearest Neighbors, Decision Tree, and Logistic Regression, were evaluated, along with ensemble methods like AdaBoost, XGBoost, and Bagging. Performance metrics, including accuracy, precision, recall, F1-score, and ROC-AUC, were used to assess the models, with the Random Forest model achieving the highest accuracy of 97.33% and an ROC-AUC score of 99.22%. This system aims to streamline the claims process and reduce fraudulent activities, ultimately benefiting insurers, policyholders, and the overall economy.