Model Predictive Control (MPC) schemes are now widely used in process industries for the control of key unit operations. Linear model predictive control schemes which make use of linear dynamic model for prediction, limit their applicability to systems which exhibit mildly nonlinear dynamics. In this paper, a state estimation based model predictive controller for nonlinear system has been proposed. The model predictive controller is designed by considering a state space model and an extended Kalman filter to predict the future behavior of the system. The efficacy of the proposed MPC scheme has been demonstrated by conducting simulation studies on the level process of a Continuously Stirred Tank Reactor (CSTR) - a MIMO system, and the real time implementation has been done in the CSTR plant to illustrate the online optimization constraint and also the advantage of MPC over conventional controller by comparison of servo-regulatory responses through ISE values. © 2013 IEEE.