As large language models (LLMs) are reshaping software development, software engineering education has not yet kept pace with these shifts in AI-driven collaborative workflows. This paper presents the design and implementation of an AI agent-based collaborative learning platform for software engineering education. Built on the MetaGPT framework, the platform simulates key development roles such as product manager, architect, engineer, and reviewer, and supports end-to-end workflows from requirements analysis to code review. On top of this foundation, we design and implement an interactive interface that enables students to observe role behaviors, intervene in agent workflows, and reflect on development outcomes. This interface supports three instructional strategies: role observation, human-in-the-loop intervention, and cross-project experience transfer, each designed to address diverse learning objectives. A formative evaluation conducted in undergraduate classrooms shows significant improvements in students’ confidence, collaboration awareness, and understanding of development workflows. These results demonstrate the feasibility and potential of the platform in bridging the gap between industrial practice and educational training.