Concurrent logic languages have been traditionally executed in a "greedy" fashion, such that computations are goal-driven. In contrast, non-strict functional programs have been traditionally executed in a "dataflow" fashion, such that computations are demand-driven. The latter method can be superior when allocation of resources such as memory is critical, which is usually the case for large, complex, and/or reactive programs. Specifically, demand-driven execution results in more efficient scheduling and improved termination properties. This paper describes a novel technique for demand­driven execution of concurrent logic language programs.