In a nutshell: Bedrock AgentCore connects agents with three layers of knowledge (enterprise, web, paid data sources) and mechanisms for productive monitoring and optimization without requiring teams to operate their own data pipelines.
Amazon introduces new capabilities for Bedrock AgentCore that equip AI agents with access to enterprise knowledge, web data, and paid information sources, along with feedback loops for continuous improvement.
Intelligent AI agents often fail in practice not due to lack of intelligence in the underlying model, but because they lack access to the necessary contextual information. A customer service agent cannot correctly answer questions about return policies if it cannot reach the corresponding document in SharePoint. A research agent delivers incomplete market reports because it cannot access current data beyond its training data. A financial advisory agent can only provide suboptimal recommendations if real-time market data from paid sources remains unavailable to it.
Amazon addresses these gaps through three new knowledge access features on the Bedrock AgentCore platform: The Managed Knowledge Base provides access to internal enterprise data from SharePoint, Google Drive, Confluence, S3, and wikis. Amazon manages the vector store, embedding and ranking models, and scalability – eliminating months of pipeline development for teams. Unlike classical vector-based search, the agentic retriever function employs multi-step planning across knowledge bases, connects related concepts across documents, and re-evaluates intermediate results to answer complex multi-part queries more precisely.
For global knowledge, Amazon introduces Web Search as a new tool integration that allows agents to research current information on regulatory changes, market developments, and competitors – while data remains within the secure AWS environment. The tool is based on Alexa and Quick Suite search infrastructure and is optimized for agentic retrieval to deliver high-quality summaries with high information density per token.
Additionally, AgentCore enables systematic monitoring of agent behavior in production – teams can determine whether their agents are improving or degrading. This closes a critical gap in continuous optimization after deployment. Together, these innovations enable teams to build more capable agents faster, implement scalable control mechanisms, and continuously improve them.
Source: aws.amazon.com · Published 17 June 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.1.