Abstract: A task scheduling method based on the improved crow search algorithm (IMCSA) is proposed, which take cloud task completion time and cost as the optimization goal. Firstly, the initial population was generated by opposition-based learning, and the preferred value was selected according to the fitness value of memory when selecting the tracked crow. That can avoid blindness. Secondly, to effectively improve the convergence speed, the position of crow was crossed with the position obtained by opposition-based learning and chose the better position during the position update process. Finally, compared with particle swarm optimization, genetic algorithm, Min_Min algorithm and CSA under the CloudSim. The experimental show that the algorithm is superior to the contrast algorithm at the task completion time and cost under different experiments.