Diagnosis of brain tumors is challenging due to their complexity and variability across individual patients. Antenatal detection of these anomalies is made by the MRI as an established imaging modality; however, manual analysis is time consuming and subject to human errors. In this research paper, we proposed a deep learning based automatic brain tumor type classification system using MRI scans. Five different CNN architectures, Custom CNN, MobileNetV2, EfficientNetB0, Xception, and EfficientNetB3 were created and evaluated. The models were built based on a public dataset with four classes i.e. Glioma, Meningioma, Pituitary Tumor, and No Tumor. EfficientNetB3 had the highest classification capability of 98.1% and was found to be better against extracting more complex spatial features than Xception (89.7%) in previous models. Application An online interactive web-based application was built to allow for easy access and practical applicability by non-experts: upon uploading an MRI image, the model automatically provides real-time predictions plus confidence scores. The system is based on the JPG and PNG formats and provides quick and user-friendly tumor detection, therefore may be usable both in clinics and remote health-care facilities. The findings in this study highlight the potential utility of deep learning and transfer learning to neuro-oncology diagnosis and offer a scalable resource for assisting radiologists for an early tumor detection and diagnosis.