The rising demand for Unmanned Aerial Vehicle (UAV) swarms in aerial surveillance, agriculture, and disaster response highlights the need for scalable testing frameworks capable of coordinating multiple drones for synchronized tasks. This paper presents M-DASTA, a Decoupled Architecture for Testing UAV Swarm Autonomy, a modular framework designed to evaluate swarm behaviors across live, virtual, and simulated environments. The architecture features independently swappable components, including controller threads that interface with Crazyflie and MAVLink-based UAVs through a unified command API. It supports both socket-based and message queue communication, and accepts diverse control inputs from keyboard commands and scripted autonomy to motion-capture-driven gestures. With protocol-agnostic integration via JSON, Protobuf, and MAVLink, the system accommodates heterogeneous agents. A key capability is native support for Live-Virtual-Constructive (LVC) testing, allowing real drones, simulators, and autonomous agents to operate under a common swarm protocol. This enables scalable, repeatable testing of swarm coordination without requiring extensive hardware. This framework aims to provide a reproducible, domain-flexible foundation for swarm autonomy research.