In order to address the constraints of obstacle shapes and global information faced by multiagents when using traditional potential methods for obstacle avoidance, as well as the tendency to get trapped in local minima in concave regions of obstacles, an obstacle avoidance method for multi-agent flocking based on behavioral mechanisms was proposed. The method assumed that the agent can only recognize other agents and obstacle boundary points within its sensing radius and effectively avoided collisions between agents and other objects through potential gradient terms. Meanwhile, a behavioral mechanism transition strategy was constructed, and by combining velocity consensus term and target guidance term, the agents’ motion behavior was classified into two modes: target tracking and obstacle circumnavigation, thus enhancing the adaptability of agents in complex environments. Finally, it was proven through contradiction that collisions between agents and other objects cannot occur during their motion, and the effectiveness of the proposed method was validated through simulation experiments.