New devices for the Internet of Things (IoT) and 5G enable monitoring and controlling of environments and objects in agriculture and other areas. Digital Twins is a growing concept connecting IoT with applications to automate agriculture, predicting crop behavior through data analysis. However, there is a conceptual gap between the digital twin concept and its application to real development. This work proposes a novel meta classes model to guide designs on digital twins in agriculture based on a bibliometric analysis to identify the current works in the area. The proposed design considers several meta classes such as communication devices, sensors, actuators, historical sensing, visualization, Human-Machine Interfaces, decisions, physical objects, and physical sectors. These meta-classes can work on a Jetson nano processor or a Raspberry Pi because they can be implemented in several languages and frameworks.