Background: Tree above ground biomass (AGB) is currently estimated by tree-level allometrical models that take into account, tree volume estimated from proxy variables of tree size (DBH) and species average wood specific gravity (WSG). These methods are common and realistic from a practical point of view. However, they do not take into account deviance from fixed allometrical trajectories and spe-cies or tree level WSG variations. Here, we present Xtrawood software that allows computation of tree AGB according to structure and WSG variations. Method: Xtrawood reconstructs tree structure and integrates WSG variations by merging tree structure and WSG data measured at different position in trees, leading to the computation of global AGB and visualization of WSG variation along tree structure. Tree structure is measured according to stem dimensions (length, diameter) and positions within tree, and encoded in Multiscale Tree Graph format (MTG). WSG data is made of radial WSG profiles (1 measure each 0,5 cm from pith to bark) sampled at different heights within whole tree. Xtrawood output are illustrated using a dataset collected on an Amazonian forest "biomass dominant species", Dicorynia guianensis Amsh., also known to exhibit substantial WSG gradients along both radial and vertical axis. 9 trees ranging from 15 to 60 cm DBH were measured by climbers. Each tree was felled and samples were collected at different positions (3 in trunk, 1 to 5 in crown) to record WSG radial profiles. Results: Xtrawood allows computation of tree volume, but also visualization of WSG variations in tree as well as inference of WSG radial profiles at different heights. Output variables are decomposed according to different tree scale and locations (axis, trunk/crown) and easy to extract. Xtrawood results will be compared to those of standard estimation method and can be used to identify positions in trees where WSG value leads to the better estimate of tree AGB. Conclusion: Xtrawood produces