A robotic system to control multiple robots to perform a task cooperatively is disclosed. A first robot determines to perform a task cooperatively with a second robot, moves independently to a first grasp position to grasp an object associated with the task, receives an indication that the second robot is prepared to perform the task cooperatively, and moves the object independently of the second robot in a leader mode along a trajectory determined by the first robot. The second robot assists the first robot in performing the task cooperatively, at least in part by moving independently to a second grasp position, grasping the object, and cooperating with the first robot to move the object, at least in part by operating in a follower mode of operation to maintain engagement with the object as the first robot moves the object along the trajectory.