Can quotas for marginalized groups in democratic institutions reduce insurgent violence? Democracy-building is a common tool of modern counterinsurgencies, yet evidence is limited about whether and how it mitigates violence. We argue that descriptive representation quotas in local government reduce insurgent violence because they create a new class of informants who increase counterinsurgents’ access to information. We find quotas are most effective where rebel institutions are more robust, and thus legible to potential informants. We estimate reservations’ effect on violence within two Naxalite insurgency-affected Indian states, Chhattisgarh and Jharkhand, with regression discontinuity and difference-in-differences de- signs. Quotas reduced violence only in Chhattisgarh, a state where Naxalites constructed sophisticated institutions. Qualitative evidence shows quotas brought potential informants closer to the state, and newly-elected officials were sometimes assassinated by insurgents for being informants. We show how institutional engineering can alter the political economy of information and trajectory of a conflict.