Teaching the Development of Software Agents in Higher Education: First Insights from a European Initiative

Abstract

Artificial Intelligence (AI) is changing software engineering by enabling autonomous software agents. This shift requires new educational approaches that prepare students to design and realise such systems. We present first experiences from an exploratory seminar in which students developed software agents using three open-source frameworks: n8n, OpenClaw, and LangChain/LangGraph. Through project-based learning, students explored agent development, prompt engineering, and the use of software models as exchange artefacts. Our observations indicate that the different frameworks offer distinct educational benefits while highlighting common challenges, including ensuring reliable software model outputs and maintaining human oversight. These preliminary results provide insights into integrating model-driven agent development into higher education and identify directions for future curricula.