Many accounts of civil war violence assume that a conflict’s master cleavage also explains the local occurrence of violence. Some scholars, however, have argued that violence is often the result of local cleavages and feuds, many of which may be unrelated to the conflict’s master cleavage. How is local violence related to the conflict’s master cleavage? Using a computational model, this paper studies an alliance mechanism proposed by Kalyvas (2006, The Logic of Violence in Civil War, Cambridge University Press), where macro-actors support local ones that fight on their behalf. While these alliances create a principal–agent problem, the model shows that they can raise the overall severity of the conflict and serve the interests of the macro-actor. However, the model also shows that these mechanisms work only under limited conditions. Alliances can increase the level of violence perpetrated in the interest of the macro-actor, but only if (a) the latter supports agents that have in the past fought along the master cleavage and if (b) this happens in rural areas. This emphasizes again the importance of the rural dimension in the study of civil war.