Conversational agents are rapidly becoming a default interface for information access and everyday work. Their societal impacts, however, are mediated by interaction: what context is retrieved, how responses are generated over multiple turns, and what interface affordances shape users’ reliance, verification, and feedback. Controlled user studies are essential for isolating these mechanisms, yet they remain difficult to run at scale. Commercial systems are fast-changing and largely black-box, while research questions often require manipulating a broad space of pipeline variants that span both agent behavior and interface design. As a result, many studies rely on bespoke prototypes that are costly to build and hard to reproduce. We present a research platform that treats conversational-agent variants as configurable study conditions. The platform supports a researcher-facing authoring interface for composing multi-stage study procedures and configuring task interfaces, alongside a participant-facing runtime that executes published study versions and logs turn-level interaction traces.