Users can define schemas with up to ten indexed keys per memory resource, including string, number, and list types, utilized to tag records during ingestion via LLM-guided extraction or explicit event metadata. Retrieval queries can filter memories by attributes like priority or date, reducing irrelevant context for agent responses.
This enhancement enables more precise and efficient memory management in AI agents, supporting structured, attribute-driven access alongside semantic similarity, ultimately improving conversational relevance and data organization across supported AWS regions.
Source & further reading
Amazon Bedrock AgentCore Memory announces metadata for long-term memory ↗Source published 2026-05-06. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.