This article sheds light on mechanisms by which online social interactions contribute to instigating far-right political violence. It presents an analysis of how violence against ethnic and religious minorities is motivated and legitimized in social media, as well as the situational conditions for such violent rhetoric. Online violent rhetoric in a Swedish public far-right social media discussion group was studied using a combination of machine-learning tools and qualitative analysis. The analysis shows that violent rhetoric primarily occurs in the context of narratives about criminals and crimes with (imagined) immigrant perpetrators and often particularly vulnerable victims, linked to a social problem definition of a corrupt and failing state as well as the alleged need to deport immigrants. The use of dehumanizing and infrahumanizing expressions both legitimizes political violence and spurs negative emotions that may increase motivation for violent action.