Amazon Bedrock AgentCore is introducing new capabilities to help agents access broader knowledge and improve continuously. These features address gaps in access to context and feedback, which have limited agents’ performance in real-world applications. The updates aim to enable agents to perform better by connecting them to organizational, web, and paid knowledge sources, as well as providing tools for monitoring and improving their performance. The changes are part of Amazon’s effort to make AI agents more effective in business environments.
The new capabilities include a Managed Knowledge Base that allows agents to access internal data from sources like SharePoint and S3 without building custom pipelines. This layer also includes an agentic retriever that goes beyond traditional retrieval methods by planning queries across knowledge bases and re-ranking results. Additionally, Web Search on AgentCore provides real-time information from the web while keeping data secure within AWS environments. These tools enable agents to reason over live web content and internal knowledge simultaneously.
According to the source, the updates are designed to help teams build more capable agents faster, govern them with scalable controls, and improve them continuously. The new features also support a trusted channel for accessing paid content, enabling both agents and content providers to benefit from a shared platform.
Source: awsml
