Amazon Bedrock AgentCore adds managed knowledge bases and web search for AI agents
Agents can now natively access internal docs and live web data without custom pipelines.
Amazon Bedrock AgentCore introduces Managed Knowledge Base and Web Search, giving agents native access to internal documents (SharePoint, Google Drive, etc.) and public web data. Features include agentic retrieval that plans queries across knowledge bases, connects related concepts, evaluates intermediate results, and re-ranks before answering—plus a multi-source grounding approach that combines public web information with Amazon’s proprietary knowledge graph. Teams can build more capable agents without custom pipelines and enforce governance within AWS security and compliance boundaries.
- Managed Knowledge Base connects agents to SharePoint, Google Drive, Confluence, S3 without custom ingestion pipelines or retrieval tuning.
- Web Search uses Amazon's search infrastructure and proprietary knowledge graph for real-time public data within AWS security boundaries.
- Agentic retrieval plans multi-part queries, evaluates intermediate results, and re-ranks across knowledge bases for more complete answers.
Why It Matters
Enables AI agents to access both internal and external knowledge, dramatically improving accuracy and usefulness in production.