Amazon Bedrock AgentCore Memory Supports Non-Conversational JSON Data for Long-Term Memory Extraction

Amazon Bedrock’s agent memory now processes structured JSON payloads directly to generate persistent memories beyond conversational inputs.

By Oliver Grant · Editorial pseudonym · 2026-08-20 · 1 min read

AI-ASSISTED REPORTING · RETROSPECTIVE ARCHIVE

Retrospective edition: 2026-08-20. Source published: 2026-08-20. Added to this archive: 2026-11-03. This briefing was prepared later from the linked announcement; it was not published on this website on the historical edition date.

Expanding beyond multi-turn conversations, AgentCore memory accepts JSON payloads up to 100 KB via its CreateEvent API. This allows agents to extract and consolidate long-term memories from varied structured data types, such as behavioral events, logs, and system events, streamlining data ingestion without conversion into conversational formats.

The memory extraction pipeline applies semantic, preference, summarization, and episodic strategies uniformly to these JSON inputs. This capability is available in all supported regions and integrates seamlessly with existing features, enhancing developers’ ability to incorporate diverse structured data into AI agent memory frameworks.

Source & further reading

Amazon Bedrock AgentCore Memory now supports extracting memories from non-conversational JSON payloads ↗

Source published 2026-08-20. 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.