Summary: This paper studies moral hazard contracts that may be renegotiated after an agent chooses an unobservable effort. Unlike in previous models, a contract here contains only one compensation scheme, and the agent has all the bargaining power in the renegotiation stage. Using a relatively weak forward-induction refinement, all equilibria are shown to be (second-best) efficient. Renegotiation occurs in every equilibrium. If the effort set is rich, the only equilibrium initial contract is a sales contract, i.e., a scheme which ``sells the project'' to the agent. This captures the idea that a party (the principal) who has an inherently weak renegotiation position will sometimes insist on a simple initial contract.