AIMS: In low-income countries, care for people with severe mental disorders (SMDs) who manage to access treatment is usually emergency-based, intermittent or narrowly biomedical. The aim of this study was to inform development of a scalable district-level mental health care plan to meet the long-term care needs of people with SMD in rural Ethiopia. METHODS: The present study was carried out as formative work for the Programme for Improving Mental health CarE which seeks to develop, implement and evaluate a district level model of integrating mental health care into primary care. Six focus group discussions and 25 in-depth interviews were conducted with service planners, primary care providers, traditional and religious healers, mental health service users, caregivers and community representatives. Framework analysis was used, with findings mapped onto the domains of the Innovative Care for Chronic Conditions (ICCC) framework. RESULTS: Three main themes were identified. (1) Focused on 'Restoring the person's life', including the need for interventions to address basic needs for food, shelter and livelihoods, as well as spiritual recovery and reintegration into society. All respondents considered this to be important, but service users gave particular emphasis to this aspect of care. (2) Engaging with families, addressed the essential role of families, their need for practical and emotional support, and the importance of equipping families to provide a therapeutic environment. (3) Delivering collaborative, long-term care, focused on enhancing accessibility to biomedical mental health care, utilising community-based health workers and volunteers as an untapped resource to support adherence and engagement with services, learning from experience of service models for chronic communicable diseases (HIV and tuberculosis) and integrating the role of traditional and religious healers alongside biomedical care. Biomedical approaches were more strongly endorsed by health workers, with traditional healers, religious leaders and service users more inclined to see medication as but one component of care. The salience of poverty to service planning was cross-cutting. CONCLUSIONS: Stakeholders prioritised interventions to meet basic needs for survival and endorsed a multi-faceted approach to promoting recovery from SMD, including social recovery. However, sole reliance on this over-stretched community to mobilise the necessary resources may not be feasible. An adapted form of the ICCC framework appeared highly applicable to planning an acceptable, feasible and sustainable model of care.