Introduction: Afghanistan has an estimated annual burden of severe acute malnutrition in children under five of over 600,000, with less then 30% accessing care. Since 2009, acute malnutrition has been part of the country's Basic Package of Health Services, which is outsourced for implementation to local partners and regulated by the Ministry of Public Health. A gradual transition to government implementation is expected to change the dynamics of health system functions and actors, with unpredictable outcomes. Building competencies in applying systems thinking by taking into account tacit knowledge may strengthen adaptive management and leadership for improving sustainable and integrated acute malnutrition interventions. We explored the dynamics of the health system with a systems lens to identify opportunities for improving sustainable and integrated acute malnutrition services to inform policies. Methods: A mixed method design involved over 70 health actors to explore the health system capacity through participatory system dynamics mapping based on rapid observation, key informant interviews, group discussions and document review. The policy analysis investigated acute malnutrition coverage. A network analysis explored involvement and influences of health actors. A framework approach appraised key health system functions and explored the level of integration of acute malnutrition. System changes over time and causal loop analyses explored system dynamics to identify leverages for improving and sustaining health outcome. Results: Key policies and strategies were in place but did not consistently recognise severe acute malnutrition as a major childhood illness. Narrow involvement of health actors missed opportunities for 'learning together' and developing sustainable and broad-based technical leadership. The health actors network showed two scale free hubs of the Public Nutrition Directorate and health workers of health facilities making the link between government and partners at and between the national/provincial and community levels. The Basic Package of Health Services left community-based nutrition underdeveloped and 40% of the hard-to-reach population uncovered. Most funding remained emergency based, and quality and ownership were limited despite major training efforts. The extent of integration was stronger at implementation level than at policy and organisational level. Behaviour-over-time, mapping the effects of financial and technical support on effective coverage, showed late but fast expansion of coverage. Causal loop analysis, building on tacit knowledge to describe the complexity of interactions and influences, found reinforcing effects from policy, competency and community involvement and dumping effects from financial and technical support on improving quality. Based on the learning from the change mechanisms, an initial theory of change identified assumptions that should be tested and refined in evaluations. Conclusion: The exploration of health system capacity and dynamics uncovered strengths and missed opportunities for sustaining integrated acute malnutrition services in Afghanistan. This study is an initial step in applying systems thinking using tacit knowledge through participative approaches to explain unpredictable behaviour and foster dialogue and 'learning together' for improving sustainable and integrated acute malnutrition services. Further research should encourage applying systems thinking to further understand dynamic complexity by opening the black box to understand why change happens, how and under what circumstances, and design effective interventions. [ABSTRACT FROM AUTHOR]