In this commentary, we take the recent introduction of methadone treatment in response to emerging problems of HIV linked to heroin addiction in Kenya as a case for reflecting on the social science of implementation science. We offer a framework of 'evidence-making intervention' which we hold as distinct from mainstream 'evidence-based intervention' approaches. Whilst accepting that interventions are shaped in their contexts, evidence-based intervention approaches tend to imagine a stable intervention object with universal effect potential. By contrast, an evidence-making intervention approach investigates how an intervention, and the knowledge which constitutes it, is made locally, through its processes of implementation. Drawing on qualitative research generated in Kenya prior to (2012-2013) and during (2014-2015) the implementation of methadone treatment, we explore the making of 'methadone promise' as a case of evidence-making intervention. We show how enactments of methadone promise make multiple methadones, through which a binary is negotiated between the narratives of methadone as hope for addiction recovery and methadone as hope for HIV prevention. Addiction recovery narratives predominate, despite methadone's incorporation into policy via its globally supported HIV prevention evidence-base. Key practices in the making of methadone promise in Kenya include its medicalization, and renaming, as 'medically assisted treatment' - or simply 'MAT' - which distance it from prior constitutions elsewhere as a drug of substitution, and the visualisation of its effects wherein unhealthy people can be seen and shown to have become well. We also show how actors seek to protect the story of methadone promise from counter narratives, including through mass media projects. We conclude that there is no single biomedical object of methadone intervening on a single biological body across contexts, and no single universe of evidence. By giving weight to local rather than outside expert knowledge, and by tracing how the meaning of intervention is made locally through its implementation, we can make visible the multiple enactments of an intervention and how these shape local ecologies of care, including in ways beyond those foreseen by an intervention's evidencing elsewhere.