AWS Expands Bedrock AgentCore with Managed Web Search and Internal Knowledge Access

AWS has introduced new capabilities for Amazon Bedrock AgentCore to connect AI agents to organizational, web, and paid knowledge. [1] The platform provides native access to three layers of knowledge, including an organizational knowledge layer called Amazon Bedrock Managed Knowledge Base. [1] This managed layer allows agents to access information stored across S3, internal wikis, Confluence, Google Drive, and SharePoint. [1]

Web Search on Amazon Bedrock AgentCore is now generally available. [2] The feature is a fully managed, Model Context Protocol (MCP)-compatible capability that agents can discover using a standard `tools/list` call. [2] It connects to an Amazon-maintained web index containing tens of billions of documents. [2] Amazon continually refreshes this index so that new content is reflected within minutes, and the privacy model ensures that search queries do not leave AWS. [2]

These updates position Amazon Bedrock AgentCore as a more comprehensive orchestration layer for enterprise AI agents. By natively integrating both internal data retrieval and a purpose-built web index of tens of billions of documents, AWS removes the engineering burden of managing third-party search APIs. This approach competes directly with ecosystem offerings from Microsoft and Google by emphasizing fully managed infrastructure and adopting the emerging Model Context Protocol (MCP) to standardize tool usage. What the sources don't address: whether the new continuous learning and feedback loop capabilities incur additional compute or storage costs for developers using the AgentCore platform.