Abnormal Cells in the Brain lead to tumor in the brain. Misdiagnosis of brain tumors will discourage efficient medical intervention and reduce patient survival chances. The life span of the patient can be increased with the early detection of tumors which is possible with CAD systems. Three class classification of Brain tumors Meningioma, Glioma, and Pituitary available from Brain database using Convolution Neural Network(CNN) is done in this paper. The images obtained from the Brain database are given as input to the convolution neural network for the classification of tumours. The CNN architecture used is simple that consists of two layers of convolution, max-pooling followed by flattening layers. The training accuracy is 98% with testing accuracy of 96% respectively with three class tumour classification.