In this research, a novel hybrid method is presented using an xLSTM-UNet architecture to provide comprehensive crop management i.e. crop recommendations and disease prediction. The approach combines U-Net feature extraction ability with temporal xLSTM modeling to improve the performance of segmentation and classification operations in agriculture related tasks. This research builds a strong framework regarding crop recommendation task. Further, the model is able to foresee possible crop diseases with the help of image processing which is done at an early stage of the crop growing process. The importance of this work is crucial as it enables farmers with useful information that improves the output of crops while reducing negative effects of farming activities on the environment. Our approach targets the challenges of disease prediction, and crop selection towards the sustainable practices of agricultural and food security. Finally, this work not only pushes the boundaries of deploying deep learning in agriculture but also offers an integrated application for the 21st century farmers towards a data driven management approach.