<p>This paper investigates the adaptive bidirectional consensus problem for multi-agent systems (MASs) subject to input saturation. First, under switching topologies, a hierarchical algorithm was introduced to address the issue of unbalanced communication structures. By means of error transformation within this hierarchical framework, the dependence on Laplacian matrix information was effectively eliminated, thereby reducing the computational complexity. Subsequently, radial basis function neural networks (RBFNNs) are employed to approximate the unknown nonlinear dynamics. A nonlinear state observer based on RBFNNs is constructed to estimate the unmeasurable states. Moreover, a flexible performance function is incorporated to handle input constraints arising from saturation, enabling the realization of prescribed transient performance in a flexible manner. An adaptive bidirectional consensus control scheme is then proposed, which guarantees finite-time tracking performance under the influence of input saturation constraints. Finally, simulation studies are conducted to validate the effectiveness of the proposed control approach.</p>