Flower strips appear to be a promising lever for promoting pest control but a careful selection of the plant species used is needed prior to implementation to avoid possible negative side effects. In the case of open field melon crops, the main pitfall would be to generate aphid and aphid-borne virus reservoirs near the crops. Combining biotests under controlled conditions and data from the literature, we assessed 18 candidate plant species, and ruled-out those posing a potential risk of hosting Aphis gossypii (melon pest and virus vector), Myzus persicae (virus vector) and/or viruses (Cucurbit aphid-borne yellows virus (CABYV), Cucumber mosaic virus (CMV), Watermelon mosaic virus (WMV) and Zucchini yellow mosaic virus (ZYMV)). Five plant species made it through the selection process: cornflower, grass pea, sainfoin, salad burnet and sweet marjoram. Flower strips sown with a mix of these five plant species were evaluated in a five-year field experiment. They displayed a flowering continuum likely to provide a food resource to natural enemies throughout the growing season. Their potential to host natural enemies was compared to those of grass strips and bare soil by monitoring generalist and specialist predators within the different field margins and melon crop. Flower margins supported significantly more of these natural enemies than grass margins and bare soil. All predator taxa analyzed responded positively to the floral resources displayed. Spiders were 3.2 times more abundant in pitfall traps placed in flower margins than in bare soil. Generalist predators and aphid specialist predators collected using a vacuum sampler were 5.5 and 9.1 times more abundant in flower margins than in bare soil, respectively. Interception traps set for weekly periods showed that coccinellid and syrphid fluxes were significantly enhanced near flower margins.