Abstract Background There is accumulating evidence for the value of collective and shared approaches to leadership. However, relatively little research has explored collective leadership in healthcare and thus, there is a lack understanding of the mechanisms that promote or inhibit the practice of collective leadership in healthcare teams. This study describes the development of an initial programme theory (IPT) to provide insight into the mechanisms underpinning the enactment of collective leadership. Methods This IPT was informed by a multiple-method data collection process. The first stage involved a realist synthesis of the literature on collective leadership interventions in healthcare settings (n = 21 studies). Next, we presented initial findings to receive feedback from a realist research peer support group. Interviews with members of teams identified as working collectively (n = 23) were then conducted and finally, we consulted with an expert panel (n = 5). Context-mechanism-outcome configurations (CMOCs) were extrapolated to build and iteratively refine the programme theory and finalise it for testing. Results Twelve CMOCs were extrapolated from these data to form the initial programme theory and seven were prioritised by the expert panel for focused testing. Contextual conditions that emerged included team training on-site, use of collaborative/co-design strategies, dedicated time for team reflection on performance, organisational and senior management support, inclusive communication and decision-making processes and strong supportive interpersonal relationships within teams. Mechanisms reported include motivation, empowerment, role clarity, feeling supported and valued and psychological safety which led to outcomes including improvements in quality and safety, staff and patient satisfaction, enhanced team working, and greater willingness to share and adopt leadership roles and responsibilities. Conclusions This study has identified preliminary support for the contexts, mechanisms and outcomes underpinning the practice of collective leadership. However, it must be noted that while they may appear linear in presentation, in reality they are independent and interlinked and generative of additional configurations. This paper contributes to the nascent literature through addressing an identified gap in knowledge by penetrating below the surface level inputs and outputs of an intervention to understand why it works or doesn’t work, and for whom it may work.