We discuss the impact of the separation of logic engines (independent logic processing units) and multi-threading on the design of coordination mechanisms for a Prolog based agent infrastructure.We advocate a combination of coroutining constructs with focus on expressiveness and a simplified, multi-threading API that ensures optimal use available parallelism.In this context, native multi-threading is made available to the application programmer as a set of high-level primitives with a declarative flavor while cooperative constructs provide efficient and predictable coordination mechanisms. As illustrations of our techniques, a parallel fold operation as well as cooperative implementations of Linda blackboards and publish/subscribe are described.