We introduce a tractable dynamic monitoring technology into a continuous-time moral hazard problem and study the optimal long-term contract between principal and agent. Monitoring adds value by allowing the principal to reduce the intensity of performance-based incentives, reducing the likelihood of costly termination. We present a novel characterization of optimal dynamic incentive provision when performance-based incentives may decline continuously to zero. Termination happens in equilibrium only if its costs are relatively low. In general, the intensity of both monitoring and performance-based compensation can be non-monotonic functions of the quality of past performance. Our results can also help explain puzzling empirical findings on the relationship between performance history and future pay-performance sensitivity and on the linkage between termination, performance, and monitoring. We also discuss implications of our model for optimal security design and endogenous financing constraints.