Greater understanding of the influences on long-term coffee productivity are needed to develop systems that are profitable, while maximizing ecosystem services and lowering negative environmental impacts. We examine a long-term experiment (15 years) established in Costa Rica in 2000 and compare intensive conventional (IC) coffee production under full sun with 19 agroforestry systems combining timber and service tree species with contrasting characteristics, with conventional and organic managements of different intensities. We assessed productivity through coffee yield and coffee morphological characteristics. IC had the highest productivity but had the highest yield bienniality; in the agroforestry systems productivity was similar for moderate conventional (MC) and intensive organic (IO) treatments (yield 5.3 vs. 5.0 t ha-1 year-1). Significantly lower yields were observed under shade than full sun, but coffee morphology was similar. Low input organic production (LO) declined to zero under the shade of the non-legume timber tree Terminalia amazonia but when legume tree species were chosen (Erythrina poepiggiana, Chloroleucon eurycyclum) LO coffee yield was not significantly different than for IO. For the first 6 years, coffee yield was higher under the shade of timber trees (Chloroleucon and Terminalia), while in the subsequent 7 years, Erythrina systems were more productive; presumably this is due to lower shade covers. If IC full sun plantations are not affordable or desired in the future, organic production is an interesting alternative with similar productivity to MC management and in LO systems incorporation of legume tree species is shown to be essential.