To manage and reason about the increasing complexity of modern network architectures, we propose Archie, a lightweight reasoning framework that enables encoding the nuances and constraints involved in deploying multiple systems—i.e., units of network functionality—in a network architecture—i.e., a collection of systems and the hardware resources they run on—in a computer-readable form. Archie serves as an “Architect-Developer Interface” between network architects, who wish to optimize objectives such as performance and cost for a given application workload, and system experts who can encode the guarantees, constraints, and costs of their systems. Given this interface, Archie allows architects to examine different configurations that optimize their objectives while checking that all their requirements are met. We explore the design space for this tool and showcase its capabilities through three case studies.