In order to meet the industry's urgent demand for portable, miniaturized and rapid detection equipment for tea quality components, a portable near-infrared tea quality detector was independently developed. This paper systematically introduced the structure, working principle, the main components and the supporting analysis software of the portable near-infrared tea detector. With the method of the partial least squares (PLS), 252 green tea samples of different grades collected from 13 different tea producing areas in China were used to establish the rapid detection models of water, tea polyphenols, amino acids, caffeine and water extracts in green tea, and the prediction effect of the models was verified. The results showed that the calibration standard deviation (SEC) and internal cross validation standard deviation (SECV) of free amino acid model were the smallest, which were 0.26 and 0.30 respectively, the calibration correlation coefficient (RC) was 0.92, andthe linear regression coefficient between the predicted value and the real value was 0.84. The SEC value and SECV value of caffeine model were 0.36 and 0.38 respectively, RC was 0.93, and the linear regression coefficient between the predicted value and the real value was 0.87. The absolute deviation between the predicted value and the real value of tea polyphenols and caffeine models was smaller than 0.50%. This study provides technical support for the popularization of portable near-infrared tea quality detector.