Amazon Bedrock AgentCore now includes web search functionality, enabling AI agents to access real-time information from the internet. This feature helps agents retrieve current data, such as stock prices or recent events, without requiring additional infrastructure. The web search capability is fully managed and integrates with the Model Context Protocol (MCP), simplifying the process for developers. It is available as a managed target or connector that connects to the AgentCore Gateway. Agents can invoke the web search tool using standard MCP calls, eliminating the need for custom API setups or result parsing.
The web search feature is powered by a purpose-built index maintained by Amazon, covering tens of billions of documents. This index is continuously updated, ensuring that the latest content is available within minutes. The system also includes a knowledge graph for high-confidence factual responses and semantic snippet extraction tailored for model context. These features enhance the accuracy and relevance of search results, making it easier for agents to provide precise and up-to-date information. Additionally, the privacy model ensures that all queries remain within AWS infrastructure, addressing data residency and security concerns for enterprises.
Amazon Bedrock AgentCore’s web search capability addresses the challenge of stale knowledge by allowing agents to access current information. Previously, teams faced obstacles such as managing third-party search APIs, parsing inconsistent result formats, and ensuring data freshness. By providing a managed solution, Amazon simplifies these processes, enabling developers to focus on building applications rather than infrastructure. The tool also supports a streamlined setup process, with minimal code required to integrate the web search function into existing workflows.
Source: awsml
