There is ongoing debate concerning whether there exists a generalizable effect of land-use change on biodiversity and consequently zoonotic disease risk. Strong data informing this debate is sparse because ecological and sampling complexities make it challenging to establish direct links between vertebrate hosts (and non-hosts), vectors, and pathogens across landscapes. However, emerging molecular methods using invertebrate-derived DNA (iDNA) can now measure species diversity and interactions from vector bloodmeals, which has the potential to improve mechanistic understanding of the effects of land-use change on zoonotic disease risk. Here, we used iDNA metabarcoding of vectors and their bloodmeals to disentangle the complex relationships between Leishmania parasites, known sandfly vectors, and potential wildlife hosts. We collected 56,775 sandflies during 3,159 trap nights at 39 forested sites across the southern Amazon ‘Arc of Deforestation’, which exemplifies global patterns of deforestation and fragmentation at the borders of tropical forest ecosystems due to agricultural expansion. We found that vector community composition was influenced by forest cover and pasture cover, and the most common vector, Nyssomia spp., was encountered less frequently in forests surrounded by pasture. Sandflies fed on a diversity of vertebrates, but the edge-loving nine-banded armadillo, Dasypus novemcinctus, was overwhelmingly the most prevalent host, followed by the greater long-nosed armadillo, Dasypus kappleri. The probability of a host being detected in sandfly bloodmeals was lower at sites with higher forest cover, which was overwhelmingly due to reduced bloodmeals arising from D. novemcinctus. Armadillos were also the most prevalent sylvatic vertebrate taxon in sandfly pools that were positive for Leishmania, further suggesting that these xenarthrans are a key host pathway for zoonotic disease transmission.