Nest architecture and demography of the non leaf-cutting fungus-growing ant species Mycocepurus goeldii and M. smithii (Attini: Formicidae) were studied in an agroforest habitat near Manaus, Brazil during the excavation of 13 nests. Both species built their nests in two different ways. The first type possessed a “tree-like” architecture, in which a vertical tunnel led downwards and lateral tunnels branched off at 90° angles from the main tunnel, with a chamber at the end of each side branch. Alternatively, other nests displayed a “necklace-like” architecture, where the main tunnel also led down vertically, but entered each chamber from the top and exited it at the bottom, resulting in an architecture where chambers appeared like pearls on a necklace. The nest systems of M. goeldii and M. smithii consisted of 1–21 or 1–15 chambers, respectively. Of 199 excavated chambers, 57 % contained fungus-gardens. Chambers not containing fungus gardens were filled with organic matter from decaying fungus gardens or earthworm feces. Only M. smithii workers deposited loose soil in abandoned chambers during the construction of new nest chambers. Workers of M. smithii constructed significantly smaller chambers than those of M. goeldii. In both species, fungus garden-containing chambers were larger than non-garden chambers and were homogenously distributed in the soil between 17 cm and 105 cm depth. Neither fungus gardens nor abandoned chambers were encountered more frequently in deeper or shallower soil strata indicating that ants of both species did not abandon shallower versus deeper chambers, or move the colony to deeper soil layers with increasing colony age. Fungus gardens were suspended from the ceiling of the subterranean chambers and originated as small mycelial tufts. Through continual addition of organic debris, the tufts first grew vertically to strands before they expanded laterally until most of the chamber volume was filled with fungus garden curtains. New garden chambers were found at depths ranging from 17 to 83 cm, suggesting that new garden chambers were not constructed in deeper soil strata with increasing colony age. The size of M. goeldii and M. smithii colonies was dependent on their age. Worker counts varied between a few individuals in recently founded colonies and 1352 workers in an adult M. goeldii colony. The ratio of worker number per fungus garden chamber was higher in M. goeldii colonies than M. smithii colonies; the M. goeldii colonies were more populous. Both species were oligogynous with a maximum of four and three queens observed in a single chamber of M. goeldii and M. smithii, respectively. The reproductive status of each queen was unknown. In both species the ratio of brood to workers was approximately 2:3. Larvae and pupae were unequally distributed throughout the nest, but were only located in chambers containing a garden. Their location in the chamber was dependent upon the developmental state of the fungus garden.