A cooperative distributed secondary/primary control paradigm for AC microgrids is proposed. This solution replaces the centralized secondary control and the primary-level droop mechanism of each inverter with three separate regulators: voltage, reactive power, and active power regulators. A sparse communication network is spanned across the microgrid to facilitate limited data exchange among inverter controllers. Each controller processes its local and neighbors' information to update its voltage magnitude and frequency (or, equivalently, phase angle) set points. A voltage estimator finds the average voltage across the microgrid, which is then compared to the rated voltage to produce the first-voltage correction term. The reactive power regulator at each inverter compares its normalized reactive power with those of its neighbors, and the difference is fed to a subsequent PI controller that generates the second-voltage correction term. The controller adds the voltage correction terms to the microgrid rated voltage (provided by the tertiary control) to generate the local voltage magnitude set point. The voltage regulators collectively adjust the average voltage of the microgrid at the rated voltage. The voltage regulators allow different set points for different bus voltages and, thus, account for the line impedance effects. Moreover, the reactive power regulators adjust the voltage to achieve proportional reactive load sharing. The third module, the active power regulator, compares the local normalized active power of each inverter with its neighbors' and uses the difference to update the frequency and, accordingly, the phase angle of that inverter. The global dynamic model of the microgrid, including distribution grid, regulator modules, and the communication network, is derived, and controller design guidelines are provided. Steady-state performance analysis shows that the proposed controller can accurately handle the global voltage regulat- on and proportional load sharing. An AC microgrid prototype is set up, where the controller performance, plug-and-play capability, and resiliency to the failure in the communication links are successfully verified.