This paper presents a novel active power flow management (PFM) method for managing multiple distributed generator (DG) units connected to medium voltage distribution networks. The method uses the artificial intelligence technique of constraint programming to autonomously manage DG real power outputs and offers flexible and network agnostic characteristics. The method is assessed using multiple scenarios on two real case study networks to examine simulated closed-loop control actions under certain thermal excursions. The test cases are explored with algorithms implemented, in software, on commercially available substation computing hardware to identify computation timescales and investigate algorithm robustness when presented with measurement error. The archival value of this paper is in the specification and evaluation of a novel application of the constraint programming technique for online control of DG in thermally constrained distribution networks.