In this work a multivariable optimal PID controller for a continuous stirred tank reactor was designed in order to obtain optimal system performance without decoupling. The tuning of this controller is obtained by using an evolutionary algorithm (MAGO). This solution is compared with classical methods for multivariable control. The results of this study showed that the controller designed rejects the effect of disturbances and follows the external reference signal efficiently. Evolutionary tuning overcomes broadly the classical methods.