This paper reports a case study of 20 university peer tutor-tutee dyads which engaged in online synchronous peer tutoring using MENTOR, a mobile application developed to support peer tutoring. Despite years of research, peer tutoring still attracts significant attention and an emerging area of research is online peer tutoring. This study aimed to contribute to research on mobile peer tutoring, which is still in its infancy stage. Underpinned by Vygotskian social-cultural learning and Wertsch's notion of mediated actions by tools, a qualitative analysis of the recorded tutees' mobile phone screen during the peer tutoring sessions was conducted. Our findings show three different types of peer tutor-tutee social cognitive interactions, with varying degrees of tutees showing agency in seeking clarifications. While most tutees demonstrated some level of agency in seeking clarifications, fewer tutees showed agency in co-annotating on the canvas space. The findings also illuminate how the participants leveraged the canvas tools provided by MENTOR to create a shared understanding and cognitive convergence.