This paper charts the frontier between strategy-proofness and collusion-proofness in efficient mechanism design, with applications to markets where pivotal firms act as hubs for collusive behavior. We model collusion via conference structures, which are collections of admissible collusive sets. A mechanism is conference-proof if no listed coalition can profitably misreport. We frame the design problem as a tension between two extremes: Groves’ schemes, which satisfy dominant strategy incentive compatibility but are proof only to singleton deviations; and Safronov (2018), which achieves full collusion-proofness but only serial Bayesian IC. Interpreting ex post budget balance as resistance to the grand coalition, we show that the AGV mechanism is weakly dominated by Cremer and Riordan (1985), and we present two novel mechanisms that strictly dominate both. Furthermore, we bridge the gap to Safronov (2018) by relaxing individual IC constraints while expanding the conference structure. Finally, we introduce a projection method to trace this incentive boundary, demonstrating that the Safronov and AGV mechanisms emerge as geometric projections of Groves’ schemes.