Real-Time Intelligence for Bedrock Agents
Amazon has launched Web Search on Amazon Bedrock AgentCore, a managed tool that enables agents to ground responses in current, cited web knowledge. By deploying the Model Context Protocol (MCP), developers can now pipe external information directly into their agents without data egress from customer’s secured AWS environment.
Infrastructure Powered by Amazon
The system routes queries through the Bedrock AgentCore Gateway. When a user inputs a natural-language prompt, the agent triggers a search that returns most relevant snippets, source URLs, titles, and publication dates. Amazon confirmed the service utilizes its internal search infrastructure—the same engine powering Alexa+, Amazon Quick, and Kiro—to synthesize multi-source grounding.
To ensure accuracy, the tool bridges standard web index results with the Amazon Knowledge Graph. This secondary layer of structured data provides verified facts, effectively insulating agents from the limitations of a model’s training data.
Securing the Enterprise Data Perimeter
For developers, initialization begins in the Bedrock AgentCore console by selecting “Connectors” as a target type within the Bedrock AgentCore Gateway. Once configured, the gateway produces a resource URL, which supports interaction via the command line interface, API calls, or the MCP Inspector.
The design prioritizes enterprise governance. By keeping retrieval queries and user prompts strictly within the AWS environment, organizations bypass the need to share sensitive information with external search API providers. This architecture allows firms to reconcile the need for real-time data with internal compliance mandates.
Early Adopters and Scientific Utility
Early testing has already yielded practical results across distinct sectors:
- Benchling: Nicholas Larus-Stone, Head of AI Agents at Benchling, noted that the tool allows scientists to ground their AI-generated hypotheses in both internal institutional data and published literature. This enables researchers to access verified, high-quality data within a secure, governed environment.
- Gen Digital: Iskander Sanchez-Rola, Senior Director of AI & Innovation at Gen Digital, reported that the tool assists in creating content ideas for the company’s "Norton Revamp" service, ensuring that recommendations remain relevant to current global events while maintaining strict data privacy standards.
Regional Rollout and Cost Structure
The service is currently live in the US East (N. Virginia) region. Amazon has opted for a usage-based pricing model, set at $7 per 1,000 search queries. There are no upfront commitments, and new AWS customers qualify for up to $200 in Free Tier credits. Users can monitor their configurations via the Bedrock AgentCore console, with technical support available through AWS re:Post or standard AWS channels.