Abstract

In this paper, we present a vision for autonomous data agent collaboration, where complex queries are no longer executed through a single centralized query plan, but instead resolved through a dynamic protocol among intelligent agents—each serving as an autonomous interface to a distinct data source. This decentralized approach eliminates the need for traditional data integration by allowing agents to reason locally while collaboratively contributing to global query results. As a core contribution toward realizing this vision, we introduce key design principles for such systems and outline the foundations of a Query Collaboration Protocol (QCP) to enable decentralized, autonomous query answering across indepen- dent and potentially noisy data sources. In our initial experiments with a prototype system, QCP-DB, we demonstrate that it can answer complex queries over heterogeneous databases with accuracy comparable to that of fully integrated centralized systems—yet without incurring the cost and complexity of data integration.