In this paper, we investigate waveform design for dynamic target tracking using a multiple-input, multiple-output (MIMO) radar system. The agile tracking application is based on our recently derived Cramer-Rao lower bound (CRLB) for estimating target position and velocity, that is represented in terms of the transmitted waveform parameters. Using the CRLB at high signal-to-noise ratio (SNR), we adaptively select the waveform parameters that minimize the predicted mean-squared error (MSE) at each time step. We demonstrate the improvement in agile tracking using numerical simulations.