Maximizing the delivery of key ecosystem services such as biological control through the management of natural enemy communities is one of the major challenges for modern agriculture. The main obstacle lies in our yet limited capacity of identifying the factors that drive the dynamics of trophic interactions within multi-species assemblages. Invertebrate generalist predators like carabid beetles are known for their dynamic feeding behaviour. Yet, at what extent different carabid species contribute to the regulation of animal and plant pests within agroecosystems is currently unknown. Here, we developed a DNA metabarcoding approach for characterizing the full diet spectrum of a community of fourteen very common carabid species inhabiting an intensively managed Western-European agroecosystem. We then investigated how diet and biological control potential within the carabid community varies with the sampling field location and the crop type (wheat vs oilseed rape). DNA metabarcoding diet analysis allowed to detect a wide variety of animal and plant taxa from carabid gut contents thus confirming their generalist feeding behaviour. The most common prey categories detected were arachnids, insects, earthworms and several plant families potentially including many weed species. Our results also show that the field location and the crop type are much stronger determinants then the species regarding carabid dietary choice: significantly more trophic links involving dipteran prey were observed in wheat, whereas more collembolan and plant prey was consumed in oilseed rape by the same carabid community. We speculate that structural differences in the habitats provided by these two crop types drive differences in resource availability cascading up the trophic chain, and we assume that specific carabid taxa could hardly be used to infer levels of ecosystem services (biological control) or disservices (e.g. intraguild predation). However, as this is the first study to report the use of DNA metabarcoding diet analysis in predatory carabid beetles we urge caution over the interpretation of our results. For instance, overall detection rates were rather low (31% of the individuals analysed tested positive for at least one prey category) most likely due to the overwhelming amplification of the carabid host DNA. Therefore, we acknowledge that more studies are required in order to confirm our observations and conclude with few recommendations for further improvements of the community-level DNA metabarcoding analysis of carabid diet.