BACKGROUND: Depression and related disorders represent a significant part of general practitioners (GPs) daily work. Implementing the evidence about what works for depression care into routine practice presents a challenge for researchers and service designers. The emerging consensus is that the transfer of efficacious interventions into routine practice is strongly linked to how well the interventions are based upon theory and take into account the contextual factors of the setting into which they are to be transferred. We set out to develop a conceptual framework to guide change and the implementation of best practice depression care in the primary care setting. METHODS: We used a mixed method, observational approach to gather data about routine depression care in a range of primary care settings via: audit of electronic health records; observation of routine clinical care; and structured, facilitated whole of organisation meetings. Audit data were summarised using simple descriptive statistics. Observational data were collected using field notes. Organisational meetings were audio taped and transcribed. All the data sets were grouped, by organisation, and considered as a whole case. Normalisation Process Theory (NPT) was identified as an analytical theory to guide the conceptual framework development. RESULTS: Five privately owned primary care organisations (general practices) and one community health centre took part over the course of 18 months. We successfully developed a conceptual framework for implementing an effective model of depression care based on the four constructs of NPT: coherence, which proposes that depression work requires the conceptualisation of boundaries of who is depressed and who is not depressed and techniques for dealing with diffuseness; cognitive participation, which proposes that depression work requires engagement with a shared set of techniques that deal with depression as a health problem; collective action, which proposes that agreement is reached about how care is organised; and reflexive monitoring, which proposes that depression work requires agreement about how depression work will be monitored at the patient and practice level. We describe how these constructs can be used to guide the design and implementation of effective depression care in a way that can take account of contextual differences. CONCLUSIONS: Ideas about what is required for an effective model and system of depression care in primary care need to be accompanied by theoretically informed frameworks that consider how these can be implemented. The conceptual framework we have presented can be used to guide organisational and system change to develop common language around each construct between policy makers, service users, professionals, and researchers. This shared understanding across groups is fundamental to the effective implementation of change in primary care for depression.