In many agroforestry systems (AFS) studies, shade was presented as being the opposite of full-sun condition. This simplification ignores light transmission through canopies or the role of diffuse light transmittance in crop light budgets. We argue here that a detailed and continuous assessment of light availability in AFS is a prerequisite to understand the impact of shade trees on the productivity of the associated crop. With this aim, we applied MAESTRA, a 3D light interception model, to a coffee AFS (CoffeeFlux Observatory) composed of two heterogeneous canopy layers, to assess the level of competition for photosynthetic active radiation absorption (aPAR) between coffee and shade trees (Erythrina poeppigiana) with a spatial resolution from plant to plot and a temporal resolution from 30 min to a whole year. Model predictions were tested against field measurements. We mapped aPAR over the coffee layer. Large and low density shade trees (9% tree cover) reduced the aPAR in coffee by 14% on a yearly average. Shade trees increased the fraction of diffuse irradiance by 20% below their crown, suggesting some positive impacts on the efficiency of coffee photosynthesis. Seasonal variations in aPAR were mainly explained by changes in coffee leaf area index with the annual coffee pruning having the strongest impact. In the actual coffee density, 35% of the incident PAR was still absorbed by the soil due to inter-row spaces; this is a large amount of underexploited energy that could be used by a cover crop. We performed prospective simulations increasing shade tree density gradually. Coffee plantation aPAR displayed a negative exponential relationship with increasing shade tree density. The photosynthesis being non-linearly related to incident light, the decrease in coffee layer photosynthesisis expected to decrease less rapidly than the decrease of aPAR. This modeling approach allows to assess the light available for individual plants as a continuous factor that can be used