Summary: When applied to groups, the revelation principle postulates a Bayesian-Nash behavior between agents. Their binding agreements are unenforceable or the principal can prevent them at no cost. We analyze instead a mechanism design problem in which the agents can communicate between themselves and collude under asymmetric information. We characterize the set of implementable collusion-proof contracts both when the principal offers anonymous and nonanonymous contracts. After having isolated the nexi and the stakes of collusion we proceed to a normative analysis, perform some comparative statics, discuss our concept of collusion-proofness, and provide some insights about transaction costs in side contracting.