An early and accurate diagnosis of brain tumors is essential due to the high mortality rate that these cause in the population. Without a doubt, identifying the type of tumor is necessary for adequate treatment and thus significantly increases the patient’s probability of survival. Magnetic resonance imaging (MRI) is crucial for diagnosing various pathologies, tumors, and brain malformations. However, manual interpretation of MRI images presents challenges for health professionals due to the complexity of brain structures and the morphological particularities of each patient. Faced with these challenges, convolutional neural networks (CNN) have demonstrated efficiency and accuracy for pattern classification tasks for various medical pathologies. A brain tumor MRI dataset is identified from the Kaggle platform. This dataset provides four brain conditions: meningioma, glioma, pituitary tumor, and no tumor. In this context, this research aims to propose a tool based on CNN to support the detection and classification of brain tumors using MRI. Then, through CRISP-DM (Cross Industry Standard Process for Data Mining), this research evaluated the main CNN and selected the optimal one to implement within a web application. In this line, the VGG16 model achieved a global accuracy of 96%, while the InceptionResNet V2, DenseNet, RestNet 50, and RestNet 101 models exhibited 95, 94, 84, and 61%, respectively. Conclusions and future work are detailed at the end of the paper.