Amazon Bedrock AgentCore Memory announces metadata for long-term memory

This article announces metadata support for Amazon Bedrock AgentCore Memory's long-term memory (LTM) records, enabling structured tagging and filtering alongside semantic search.

• Define up to ten indexed metadata keys per memory resource (STRING, NUMBER, STRING_LIST types)
• Attach metadata at ingestion time or auto-infer via LLM extraction instructions
• Filter retrieval results using structured attributes like ticket number, priority, or date
• Improves agent response accuracy by eliminating irrelevant context
• Available today in all AWS Regions supporting Amazon Bedrock AgentCore Memory

This enhancement enables agents to organize and retrieve memories more efficiently using both semantic and structured metadata filtering.