Why do some communities experience more violence against civilians than others? This study argues that civilian communities embracing relationships that bridge salient group divides and norms of non-violence, will commit to actions of protection across those divides, which will in turn limit possibilities for armed actors to commit violence against civilians. On the other hand, communities that do not embrace bridging relationships and instead promote more violent norms, will commit to denunciation of other civilians during war. This will in turn create more opportunities for armed actors to commit violence against civilians. Through interview-based field research I test this theory on two communities in Kosovo. I compare the town of Prizren, largely spared from violence against civilians during the war of 1998-1999, to the town of Gjakova, which experienced higher levels of violence against civilians. The results show support for the theoretical argument. However, I cannot fully account for alternative explanations to the difference in violence. Research on civilian agency during the Kosovo war has been severely lacking, an empirical gap which I aim to partly fill through this research.