This paper is concerned with the consensus of multi-agent systems with stochastic controller gain variation.Firstly, under the hypothesis that the connected network topology is fixed, undirected and linked, a closed-loop system is obtained that captures the stochastic controller gain variation.By using the Lyapunov stability theory, linear matrix inequality technique and robust control approach, a new sufficient condition is obtained such that consensus error dynamic system is asymptotically stable.Then, after some matrix manipulation, the controller gains are determined by solving a set of linear matrix inequalities.Finally, the effectiveness of the proposed design algorithm is verified by a numerical simulation.