Crop disease treatment is vital for improving agricultural production and crop yields. The early prevention and treatment of the disease is very helpful to reduce crop damages. However, traditional crop disease treatment is much costly and time consuming to counsel with an agriculturist. In Bangladesh, most of the farmers are unaware of pest control and disease treatment. In order to overcome this problem, KrishokBot is deployed. It is a smart agent that makes remote interaction with farmers to provide pest and disease related solution using natural language processing. KrishokBot is a Machine Learning based virtual assistant that can respond to simple questions concerning pests and disease that affect rice production via Bengali language. The datasets have been collected from various Bangladeshi agriculture-based websites to train the KrishokBot, which includes categories, patterns, and responses. A deep neural network has been used to determine which category the user’s message belongs to, and then a response is generated. For this, some tools are used such as Natural Language Tool Kit, Keras API, Tensorflow, Android SDK, Android Volley, Heroku, etc. This proposed idea offers great potential for excellent performance with approximately 85 percent accuracy, where user Interface has been developed by android application with both audio and text-based features to provide better interaction. The results prove that the bot is reliable for guiding the treatment of crop disease.