This paper proposes a two-way amplify-and-forward (AF) multiple-input multiple-output (MIMO) beamforming relay network for the benefit of determining a set of relay amplifying matrices and transmit/receive beamforming vectors by imposing constraints on the transmit power of the sources and relay based on the minimum mean square error (MMSE) criterion. It is proven that minimizing the mean square error (MSE) in the two-way AF MIMO beamforming relay system is equivalent to maximizing the signal-to-noise ratio (SNR). Additionally, various analysis, such as SNR and MMSE cost function, will be investigated. Finally, the iterative algorithm is proposed to solve the desired optimization problem.