Biology is dynamic: the location, shape, and function of molecules and cells change with time. When studying biological systems, it is critical to adapt data collection and analysis to fit their current states. However, the lack of real-time interactions in traditional microscopy makes it impossible to guide the experiments and adapt to the biological events. In this thesis, we introduce closed-loop microscopy (CLM) approaches that address the current shortcomings by providing real-time interactions between acquisition and analysis. CLM is implemented in an event-driven way; acquisition events notify the downstream analysis, resulting in feedback that triggers real-time actions. CLM is particularly suited for long experiments that study rare biological events; experiments in which adapting to the real-time changes increases the probability of success. We demonstrated examples in which CLM reduced sample size variation across trials, achieved five times higher throughput in fluorescent protein characterization, and enabled the study of rotavirus at low multiplicities of infection.