Skip to content
0MESubscribe
AI17 Nov 20262 min read

Amazon Bedrock links AgentCore memory directly to agentic knowledge base searches

Amazon Bedrock now connects its AgentCore memory feature with Managed Knowledge Base, letting AI agents pull long-term user context and document data in one API call.

Listen to this storyAbout 3 min, read by your device
Picture made with AI

Amazon Bedrock has updated its Managed Knowledge Base to natively support persistent memory. By integrating Amazon Bedrock AgentCore memory into the AgenticRetrieveStream APIAPIA way for one program to ask another program for data or a service. Open in glossary, the service allows AI agents to automatically recall past user facts and search company documents in a single operation.

Automating context and long-term memory

Knowledge-based assistants often fail to connect previous conversations with current search requests. Users expect an assistant to recall their role, location, or past interactions without re-explaining them. Previously, developers had to build custom databases to store chat history, write code to pull durable facts, and then insert those facts into new queries.

With the new opt-in memory configuration, Bedrock handles these steps directly. When an application sends a query, the service restores the dialogue history from short-term memory. An internal planner determines whether user-specific details from long-term memory will sharpen the search. It retrieves those facts, constructs sub-queries for the knowledge base, provides a grounded response with citations, and writes the latest turn back to memory.

Multi-tenant security and access controls

The architecture allows multiple users to share a single knowledge base and memory resource without data leaking across accounts. Systems can isolate data by tying verified user identities to access control lists and memory namespaces.

In a reference setup using Amazon Cognito and Amazon S3, an authenticated tokenTokenA small piece of text, about three quarters of a word, that an AI model reads or writes. Open in glossary provides both the user ID and email:

  • The verified email checks document-level access rules, ensuring users only retrieve files they are cleared to read.
  • The user ID scopes both short-term session recall and long-term memory records.
  • AWS Identity and Access Management policies enforce memory isolation even if application code fails.

Because the platform automatically manages retention, encryptionEncryptionA security method that turns readable information into an encoded secret code to protect data. Open in glossary, and extraction, developers no longer need to maintain separate conversation tables for multi-tenant enterprise assistants.

How does this story land?

One tap per reader.

Quick poll

Should cloud platforms manage AI memory natively instead of relying on custom databases?

One tap. No sign-up.

Ask this story

Still have a question?

Answers use only what is in this story. They are written by AI and can be wrong.

Sources

  1. Artificial Intelligence, Build memory-aware AI agents with Agentic Retrieve on Amazon Bedrock Managed Knowledge Base
Companies in this storyAmazonWords explainedAPITokenEncryption
Up next in AIWhy 88% of AI pilots fail and how multi-agent patterns fix coordination risks2 min read

Keep reading

All AI stories