ECHO Offers Auditable, Cognitively Inspired Memory for AI Agents.

Long-horizon AI agents require sophisticated memory systems that can not only recall relevant past experiences but also manage revisions and provide clear audit trails for their decisions. Existing memory solutions often lack these capabilities, making it difficult to understand or debug agent behavior over extended periods. ECHO (Embodied Context and History Orchestration) presents a novel memory architecture and prototype service designed to address these needs. Drawing functional inspiration from human cognitive processes such as episodic encoding, consolidation, and executive control, ECHO aims to create a memory plane that is both effective and auditable. The system's empirical analysis focuses on its ability to retrieve and construct context efficiently. While development runs show high hit rates and recall on various benchmarks, a post-hoc audit revealed that query-expansion rules contained source-specific phrases, indicating that some retrieval scores are descriptive development measurements rather than independent confirmations of its full robustness. Nevertheless, ECHO represents a significant step towards more transparent and reliable long-horizon agent memory.