A costly 'facility' has a monopoly on the ability to coerce transfers and verify all private information. If invoked, the facility reads instructions recorded ex ante, and carries out the contingent transfers among agents, charging agents for the cost. Agents agree ex ante to a set of recorded instructions to the facility, and then play a sequential game without commitment. A basic two-agent insurance environment serves as an application throughout. When both agents have full information the costs and limitations of the facility constrain the set of attainable allocations, even though the facility is never invoked in equilibrium. When there is private information, the model can be viewed as a reformulation of the standard costly-auditing model, but the incentive constraints are significantly more severe. Pure strategy optimal contracts are debt contracts, as in Townsend (1979), but mixed strategy optimal contracts cannot be ruled out in general. An extension shows that if costs vary with the realized state in a particular way, debt contracts as in Williamson (1987) can be optimal, even allowing for mixed strategies.