Summary: We analyze renegotiation in a hidden action principal-agent model. Contract renegotiation offers are made by the agent. A refinement is imposed on the principal's beliefs: if precisely one action is optimal with respect to both the principal's and the agent's contracts, the principal believes that action has been taken. With the refinement imposed, perfect-Bayesian equilibrium allocations are identical to the second best in the classical principal-agent model without renegotiation. When renegotiation is led by the agent and when equilibria satisfy the refinement, equilibrium allocations are ex ante efficient.