This paper proposes a formation controller for multi-agent systems focusing on the obstacle avoidance and input constraints. In our proposed method, the limitation of translation and angular velocities is adaptably tuned depending on the existence of obstacles, such that the velocities can increase when the robots potentially hit obstacles. Moreover, a potential function of the input velocities that easily satisfies velocity constraints is designed. We also design a formation controller that realizes obstacle avoidance while satisfying the velocity constraints. Mathematical analysis shows the stability of the formation of robots controlled by the proposed model. Simulation results also demonstrate the effectiveness of the proposed method.