We propose a distributed control law to increase the robustness of a multi-agent network to node failure or removal. More precisely, our approach is able to maintain biconnectivity of an initially biconnected graph. Remarkably, if the graph is not initially biconnected or if the property is lost after a node removal or failure, our approach is also able to render the graph biconnected after a time instant which can be tuned by the user. The proposed control algorithm is completely distributed using only locally available information and requires the estimation of a single global parameter akin to existing connectivity-maintenance algorithms in the literature. Numerical simulations illustrate the effectiveness of our approach.