Collaboration scripts are widely employed in online collaborative learning to enhance student engagement and facilitate collaboration. However, the optimal level of scripting remains a subject of debate. This study aims to address this issue by designing and developing different types of collaborative scripts implemented through conversational agents and supported by WeChat. Utilizing interventional studies, we investigate the effects of these different collaboration scripts on student engagement during online collaborative learning. A total of 54 college students participates in the study, divided into six adaptable scripts teams, six maximal script team, and six minimal script team, with each team consisting of three students. Both quantitative and qualitative data are collected and meticulously analyzed. The results reveal that the maximal collaboration script significantly enhances cognitive interactions, whereas the minimal collaboration script fosters high-quality cognitive engagement. In terms of socio-emotional engagement, the adaptable collaboration script effectively promotes positive socio-emotional engagement, while the maximal collaboration script facilitates greater socio-emotional interactions. Furthermore, thematic analysis demonstrates that all three types of collaboration scripts support student engagement by providing time reminders, facilitating planning, clarifying ideas, and promoting task reflection. These findings have important implications for improving group learning engagement in online collaborative learning environments.