Background Theories and meta-analyses have elucidated individual-level mechanisms of action in behaviour change interventions. Although group-based interventions are commonly used to support health-related behaviour change, such interventions rarely consider theory and research (e.g. from social psychology) on how group-level mechanisms can also influence personal change. Objectives The aim was to enhance understanding of mechanisms of action in group-based behaviour change interventions. The objectives were to (1) develop a potentially generalisable framework of change processes in groups, (2) test the framework by analysing group session recordings to identify examples illustrating group processes and facilitation techniques and (3) explore links between group-level mechanisms and outcomes. Data sources In this mixed-methods study, the research team reviewed literature, conducted consultations and analysed secondary data (i.e. delivery materials and 46 audio-recordings of group sessions) from three group-based weight loss interventions targeting diet and physical activity: ‘Living Well Taking Control’ (LWTC), ‘Skills for weight loss Maintenance’ and ‘Waste the Waist’. Quantitative LWTC programme data on participant characteristics, attendance and outcomes (primarily weight loss) were also used. Methods Objectives were addressed in three stages. In stage 1, a framework of change processes in groups was developed by reviewing literature on groups (including theories, taxonomies of types of change techniques, qualitative studies and measures of group processes), analysing transcripts of 10 diverse group sessions and consulting with four group participants, four facilitators and 31 researchers. In stage 2, the framework was applied in analysing 28 further group sessions. In stage 3, group-level descriptive analyses of available quantitative data from 67 groups and in-depth qualitative analyses of two groups for which comprehensive quantitative and qualitative data were available were conducted to illustrate mixed-methods approaches for exploring links between group processes and outcomes. Results Stage 1 resulted in development of the ‘Mechanisms of Action in Group-based Interventions’ (MAGI) framework and definitions, encompassing group intervention design features, facilitation techniques, group dynamic and development processes, interpersonal change processes, selective intrapersonal change processes operating in groups, and contextual factors. In stage 2, a coding schema was developed, refined and applied to identify examples of framework components in group sessions, confirming the content validity of the framework for weight loss interventions. Stage 3 demonstrated considerable variability in group characteristics and outcomes and illustrated how the framework could be applied in integrating group-level qualitative and quantitative data to generate and test hypotheses about links between group mechanisms and outcomes (e.g. to identify features of more or less successful groups). Limitations The framework and examples were primarily derived from research on weight loss interventions, and may require adaptations/additions to ensure applicability to other types of groups. The mixed-methods analyses were limited by the availability and quality of the secondary data. Conclusions This study identified, defined, categorised into a framework and provided examples of group-level mechanisms that may influence behaviour change. Future work The framework and mixed-methods approaches developed provide a resource for designers, facilitators and evaluators to underpin future research on, and delivery of, group-based interventions. Funding This project was funded by the Efficacy and Mechanism Evaluation programme, a Medical Research Council and National Institute for Health Research partnership.