•We review the use of agent-based simulation (ABS) in agrifood supply chain research.•Findings include common ABS model structures and modelling approaches.•It complements existing reviews on the use of ABS in related domains.•Bibliographic mapping is used to highlight topics for future research. Agri-food supply chains (ASC) are an important application domain for Operational Research/Management Science. In particular, the use of agent-based simulation (ABS) has increased in ASC research in recent years. This paper reviews existing ASC research that use ABS. The review begins by analysing the characteristics of the models and modelling reported in the literature. It illustrates that existing modelling research features extensive use of: single echelon supply chains; cases from high and middle income countries; unprocessed food products, empirical (as opposed to hypothetical) data; decision-making related to production planning and investment; and the use of black box validation. The second part of the review uses bibliographic mapping to analyse areas in ASC research which are yet to be addressed using ABS. We find that areas such as collaboration and competition, buyer–seller relationships, and service are under-researched. In addition, key actors in ASC such as food processors, supermarkets and retailers have not been included in the ABS models reported. Furthermore, these models have yet to incorporate important supply chain management theories such as Transaction Cost Economics and Resource-Based View as part of their design.