Aims To show how the inclusion of agent-based modelling improved the integration of ethno-epidemiological data in a study of psychostimulant use and related harms among young Australians. Methods Agent-based modelling, ethnographic fieldwork, in-depth interviews and epidemiological surveys. Setting Melbourne, Perth and Sydney, Australia. Participants Club drug users in Melbourne, recreational drug users in Perth and street-based injecting drug users in Sydney. Participants were aged 18-30 years and reported monthly or more frequent psychostimulant use. Findings Agent-based modelling provided a specific focus for structured discussion about integrating ethnographic and epidemiological methods and data. The modelling process was underpinned by collective and incremental design principles, and produced 'SimAmph', a data-driven model of social and environmental agents and the relationships between them. Using SimAmph, we were able to test the probable impact of ecstasy pill-testing on the prevalence of harms - a potentially important tool for policy development. The study also navigated a range of challenges, including the need to manage epistemological differences, changes in the collective design process and modelling focus, the differences between injecting and non-injecting samples and concerns over the dissemination of modelling outcomes. Conclusions Agent-based modelling was used to integrate ethno-epidemiological data on psychostimulant use, and to test the probable impact of a specific intervention on the prevalence of drug-related harms. It also established a framework for collaboration between research disciplines that emphasizes the synthesis of diverse data types in order to generate new knowledge relevant to the reduction of drug-related harms.