Mental health crises cause significant disruption to individuals and families and can be life-threatening. The large number of community crisis services operating in an inter-agency landscape complicates access to help. It is unclear which underpinning mechanisms of crisis care work, for whom and in which circumstances. The aim was to identify mechanisms to explain how, for whom and in what circumstances adult community crisis services work. The objectives were to develop, test and synthesise programme theories via (1) stakeholder expertise and current evidence; (2) a context, intervention, mechanism and outcome framework; (3) consultation with experts; (4) development of pen portraits; (5) synthesis and refinement of programme theories, including mid-range theory; and (6) identification and dissemination of mechanisms needed to trigger desired context-specific crisis outcomes. This study is a realist evidence synthesis, comprising (1) identification of initial programme theories; (2) prioritisation, testing and refinement of programme theories; (3) focused realist reviews of prioritised initial programme theories; and (4) synthesis to mid-range theory. The main outcome was to explain context, mechanisms and outcomes in adult community mental health crisis care. Data were sourced via academic and grey literature searches, expert stakeholder group consultations and 20 individual realist interviews with experts. A realist evidence synthesis with primary data was conducted to test and refine three initial programme theories: (1) urgent and accessible crisis care, (2) compassionate and therapeutic crisis care and (3) inter-agency working. Community crisis services operate best within an inter-agency system. This requires compassionate leadership and shared values that enable staff to be supported; retain their compassion; and, in turn, facilitate compassionate interventions for people in crisis. The complex interface between agencies is best managed through greater clarity at the boundaries of services, making referral and transition seamless and timely. This would facilitate ease of access and guaranteed responses that are trusted by the communities they serve. Strengths include the identification of mechanisms for effective inter-agency community crisis care and meaningful stakeholder consultation that grounded the theories in real-life experience. Limitations include the evidence being heavily weighted towards England and the review scope excluding full analysis of ethnic and cultural diversity. Multiple interpretations of crises and diverse population needs present challenges for improving the complex pathways to help in a crisis. Inter-agency working requires clear policy guidance with local commissioning. Seamless transitions between services generate trust through guaranteed responses and ease of navigation. This is best achieved where there is inter-agency affiliation that supports co-production. Compassionate leaders engender staff trust, and outcomes for people in crisis improve when staff are supported to retain their compassion. Further work might explore inter-agency models of crisis delivery, particularly in rural communities. Future work could focus on evaluating outcomes across crisis care provider agencies and include evaluation of individual, as well as service-level, outcomes. The implementation and effect of mental health triage could be explored further, including via telehealth. Barriers to access for marginalised populations warrant a specific focus in future research. The study is registered as PROSPERO CRD42019141680. This project was funded by the National Institute for Health and Care Research (NIHR) Health and Social Care Delivery Research programme and will be published in full in Health and Social Care Delivery Research; Vol. 11, No. 15. See the NIHR Journals Library website for further project information.