Task scheduling is of great significance to shorten performing time and minimize the cost for computational grid. A grid task schedule algorithm is presented in this paper, which is based on a constraint satisfaction neural network. The constraint satisfaction means to remove the violations for sequence and resource constraints during scheduling subtasks for grid environment. The data-transferring costs among subtasks are also considered in our task scheduling. The simulation in this paper has shown that the task schedule algorithm is efficient with respect to the quality of solutions and the solving speed.