The production and management of violence is a central element to the order-building process during civil wars. I theorize that the conditions which produce violence between and within communities are the product of dynamic interactions between combatants and civilian communities. Combatants either directly administer control of territory or delegate to local allies, options that I term direct and indirect control. Community interactions with combatants are shaped by their social cohesion, influencing their capacity for collective action and ability to resolve internal disputes. The interaction between modes of combatant control and community cohesion produces what I term “security orders,” reflecting the ability, or inability, of combatants and communities to manage violence within communities. Different forms of security orders produce varying predicted relative levels of communal violence. The theory is applied to the U.S. Civil War, including a new dataset of violence from the U.S. Civil War collected through extensive archival research, and the African Union intervention in contemporary Somalia. I find strong empirical support for the theoretical argument that indirect forms of control over low cohesion communities will produce the highest levels of communal violence, while direct forms of control over similar communities can suppress violence. High cohesion communities demonstrate the capacity to maintain relative peace under multiple forms of control. I highlight the importance of communal violence in determining local order during wartime and the central role identity plays in driving violence between groups. Additionally, I provide a unique model of territorial control, capturing heterogeneity in combatant behavior and the effects of local alliances. Finally, the U.S. Civil War and Somalia are prominent examples of state-building through conflict, offering crucial cases to understand the micro-foundations of order-building.