Abstract

This study investigates the role of large language model (LLM) agents in enhancing student learning in computing education. Through a systematic review of the literature, we analyzed 46 peer-reviewed publications published between 2020 and 2025, retrieved from three academic databases, to understand how current LLM agents support learning and the associated challenges. We categorized the contributions of the LLM agent into five areas: personalized support, accessible assistance, cognitive scaffolding, skill enhancement, and student engagement. Although the literature indicates promising pedagogical benefits, it also reveals recurring concerns, including dependency risks, reliability concerns, technical barriers, and ethical issues. Based on these findings, we propose a framework for the responsible integration of LLM agents into computing curricula, contributing to the advancement of AI in education.