In Central Cameroon cocoa is mainly produced by household farming systems based on complex associations between cocoa and companion trees. Setup either on native/remnant forest or savannah, these agroforestry systems (AFS) are managed according their geographical position and local pedoclimatic conditions. In this paper, we investigated the effects of local management strategies on carbon (C) storage of live trees in three different cocoa production zones of Central Cameroon. In the 58 fields studied, 8,996 cocoa trees and 1,258 companions were surveyed. Tree sampling was non-destructive and to estimate C storage we used allometric models for above- and belowground biomasses. We measured abundance, height, diameter at breast height and determined species of companion trees. We distinguished between four cocoa plantation age categories (immature, young, mature and senescent) and three preceding systems (forest, forest gallery and savannah). We surveyed farmers' use of each associated tree, allocated it to a functional category and asked if it had been introduced or conserved. Total C content of live trees was on average close to 70 t ha?1. We found that it mostly relied on associated trees-cocoa trees contribution being ac. 2-12 % of live trees total C. The level of contribution to C storage of companions from different use categories differed between sites-trees producing food had contributed most in Bokito and Obala while trees used for shading or fertility contributed most in Ngomedzap. Dynamics of C storage in live trees was found to be independent from cocoa trees growth and age. When aging, AFS continuously lost companion trees and especially conserved ones putatively because of farmers' selective logging. Yet, AFS apparently maintained equivalent C storage abilities with time. Hence, even if cocoa do not contribute significantly to C storage in our study, the systems into which they are included are able to significantly store C and may also contribute to othe