Accurate egg production forecasting is essential for resource optimization, farm profitability, and sustainable poultry management. StackEgg, a stacked ensemble approach, is introduced in this study that combines a set of tree-based regressors with a linear meta-model to optimally capture subtle, nonlinear relationships in egg production data. Egg production data of 60 commercial layer farms in three districts of Bangladesh and covering birds from 1 to 90 weeks of age were used to train and test the model. Random Forest, XGBoost, and CatBoost are used as base learners, whose results are aggregated by a Linear Regression meta-learner, which is trained using 5-fold cross-validation. Experiment results indicate that StackEgg achieved the lowest MAE of 0.56, RMSE of 5.61, and MAPE of 0.99% on independent test sets and outperformed regression methods in general consistently. In addition, SHAP and LIME were used to enhance interpretability, which confirmed that age, body weight, and feed intake are the primary drivers of egg production. The findings highlight StackEgg's worth as a stable and transparent decision-support system for precision poultry farming.