This work addresses the issues of robust consensus in descriptor multi-agent systems. The interaction topology among agents is considered as a directed graph and the proposed distributed consensus protocol is output feedback-based. The robust consensus control problem is reformulated through a unified framework model into a robust stability problem for a single agent. The controller gain parameters are obtained by solving a set of linear matrix equality constraints and strict linear matrix inequality constraints. These constraints are derived from the bounded real lemma. The effectiveness of the proposed methodology is illustrated through a numerical example.