The body temperature is one of the important parameters for cow health status and also is one of great significance reference for disease diagnosis and treatment. The data collection of cow body temperature by manual cant be realtiming and sometimes has errors so that brings a lot of difficulties to management. In view of development of computer technology and applications in the agriculture, based on the current mainstream of WiFi, a realtime monitoring system was applied to inspect the variation of cow body temperature. The digital temperature sensor DS18B20 was used to collect cow’s aural temperature, communicated with ESP-WROOM module that was integrated Tensilica L106, 32bit microcontroller (MCU), a 16bit RSIC with the complete system and selfcontained WiFi networking capabilities through onewire connector. At the first time ESP-WROOM module must be configured by SmartConfig protocol with APs (Access point) parameters which aiming to be slinked with AP. Once the connection was successfully established, the temperature data will be sent to MCU. After received, MCU packs data and parameters send to server through TCP protocol. And the data collected were stored in MySQL database respectively in different cows. The soft system programmed with Java language can achieve realtime acquisition, storage, display and inquiry of cows temperature data. The asynchronous communication models in view of Java NIO class were used to solve the masses of service requests delivered and the request congestion excluded. The results of experiment showed that errors of the proposed device were ±0.2℃ with the range of 30~50℃, the proper transfer distance of sensor node was 80m, and the loss rate of network passage was 4.42%. The power consumption was decreased with collecting time and sleep mode. It is proved that the characteristic of lowpower, small bulk, high precision of sample can be provided and applied practically.