We developed an automatic triage tool for emergency department settings using AutoScore, a methodology that combines the non-linear modeling capability of machine learning with the interpretability of logistic regression. The tool generates point-based clinical scores that are both accurate and transparent. Applied to real-world emergency department data, the model outperformed existing triage scores in predicting patient severity. Its simplicity and structured output allow for seamless integration into electronic medical record (EMR) systems, supporting real-time, interpretable decision-making in clinical workflows.