The direct agglutination test (DAT) was evaluated for serodiagnosis of visceral leishmaniasis (VL) in human and canids (dogs and foxes Cerdocyon thous). The results were compared with those of the immunofluorescent antibody assay (IFAT) and enzyme-linked immunosorbent assay (ELISA). The sera used were from: humans (303): confirmed VL (16), suspected VL (65), other conditions (102), negative controls (15) and individuals from an endemic area (105); dogs (82): from an endemic area (68), Salvaterra/Marajó/Pará (21 of which were parasitologically positive), and negative controls (14), from Belém; foxes (9): caught on Marajó Island. Antigens for DAT were prepared from promastigots of L. (L.) donovani, L. (L.) chagasi. Antigens used in ELISA and IFAT were prepared from promastigotes (soluble antigen) and amastigotes respectively of L. (L.) chagasi. In humans, the specificity and sensitivity of DAT using L. (L.) donovani were high (98.4% and 100% respectively) and comparable to that of IFAT (97.5% and 100%). ELISA was less specific (84.8%) although similarly sensitive (100%). In dogs, DAT was more specific using L. (L.) donovani as antigen than using L. (L.) chagasi. However, both DAT and ELISA were less sensitive (both 71.4%) than IFAT (100%). This difference was reflected in the results from endemic dogs, 87% of which were positive by IFAT but only 54% by ELISA and 49% by DAT. Similarly, all 9 fox sera were positive by IFAT, 7 of 9 (78%) by ELISA but none by DAT. In conclusion, DAT using L. (L.) donovani antigen can provide a useful test for human VL; utilization on a large scale would be possible with a suitable reference laboratory to monitor antigen quality. However, DAT appears less useful for canine studies, as it was less sensitive than ELISA and especially IFAT in detecting canine infection.