The potential for tree components of coffee agroforestry systems to provide ecosystem services is widely recognized. Management practices are a key factor in the amount and quality of ecosystem services pro- vided. There is relatively abundant information on ecosystem services provision within agroforestry sys- tems, but comparatively scant information regarding how coffee farmers manage their plantations, the factors influencing their farming practices and the extent to which farmers’ local knowledge – as opposed to global scientific understanding – underpins management decisions. Policymakers and scientists too frequently design development programs and projects in the coffee sector. On occasion technicians are included in the design process, but farmers and their knowledge are always excluded. This research explores farmers’ knowledge regarding how trees affect coffee productivity and ecosystem services in Costa Rica. Farmers’ knowledge on the effects of trees on coffee productivity was compared with that of other knowledge sources: coffee processors, technicians and scientists. Farmers were shown to have detailed knowledge regarding ecosystem services that their coffee agroforestry systems provide as well as on the interactions between trees and coffee productivity. When asked on the services that trees pro- vide, farmers classified trees according to water protection, soil formation, or contribution to biodiversity conservation. These classifications were related to tree attributes such as leaf size, biomass production or root abundance. Comparison of coffee productivity knowledge from different knowledge sources revealed considerable complementarity and little contradiction. The effects of shade trees on biophysical conditions and their interactions with coffee productivity were well understood by farmers. They recorded and classified shade trees as ‘fresh’ (suitable for integra- tion with coffee) or ‘hot’ (unsuitable) based on their leaf texture and size, foliage density, crown shape, and root system attributes. The fresh/hot classification significantly related to positive/negative provision of services. This classification was widely used by farmers, and unknown by coffee technicians. Detailed local knowledge included several different topics, such as the role of trees in soil formation and in abundance of pollinators. Farmers were also aware of the influence of these ecosystem services on crop productivity. Generally, management decisions were made to maintain coffee productivity rather than ecosystem services. Based on these results, it is suggested that technical interventions addressing the improvement of coffee plantations are more likely to be successful if they take into account not only the scientific information on agroforestry interactions but also the knowledge possessed by farmers. Lack of comprehension of local coffee knowledge could be expected to reduce the success of development pro- grams and projects aimed at improving productivity or other ecosystem services.