A mobile multimodal classification system for diseases of agricultural plants is presented. A MobileNetV2-based neural network architecture has been developed that integrates visual data with agronomic metadata. An Android application has been created using TensorFlow Lite for real-time inference. Validation on the cucumber diseases dataset showed an accuracy of 99.4%, which is 4-5% higher than unimodal analogues. The system provides rapid diagnostics in the field without an Internet connection.