The accuracy of forecasting the price of agricultural products may help farmers and agricultural product entrepreneurs in their productions planning well. Today, agricultural product price forecasting using econometric methods are always done by using general statistical analysis programs which are difficult, and may need some expert in statistics. Propose of this thesis is to design and develop the automatic agricultural products prices forecasting system using econometric methods. The system will be able to help farmers and the forecasters who are not expert in statistics to forecast the agricultural product prices easier. The forecasting system is designed by using three individual forecasting methods: decomposition method, causal method, and exponential smoothing methods. All three forecasting methods have automatic parameter optimization. Three combinations of the above three forecast methods are used to increase forecasting accuracy. There are simple averages, weights inversely proportionate to sum ofsquared errors, and weights determined by regression analysis. Recursive cross-validation is used to select the appropriate forecast combinations. The test of the developed system was done by using data of three agricultural products: rice, cassava and rubber, from related government departments, for twelve periods forecast. The evaluation of tested result using mean absolute percentage error to measure accuracy shows that accuracy of the forecast system is good to very good.