Urban violence is a major preoccupation of policymakers, planners and development practitioners in cities around the world. Public authorities routinely seek to contain such violence through repression, as well as through its exportation to and containment at the periphery of metropolitan centres. Yet, urban violence is a highly heterogeneous phenomenon and not amenable to reified diagnosis and coercive intervention. Muscular state-led responses tend to overlook and conceal the underlying factors shaping the emergence of urban violence, as well as the motivations and means of so-called violence entrepreneurs. This is very obviously the case of urban gangs in Central America, which are regularly labelled a ‘new urban insurgency’ threatening the integrity of governments and public order. This article considers both the shape and character of Central American gang violence and attempts to reduce it, highlighting the complex relationship between these two phenomena. We advance a threefold approach to measuring the effectiveness of interventions, focusing in turn on discursive, practical and outcome-based criteria. In this way, the article demonstrates how, contrary to their reported success in diminishing gang violence, repressive first-generation approaches have tended instead to radicalize gangs, potentially pushing them towards more organized forms of criminality. Moreover, although credited with some modest successes, more preventive second-generation interventions seem to have yielded more rhetorical advances than meaningful reductions in gang violence.