AWS launches Web Search on Amazon Bedrock AgentCore with proprietary index at $7 per 1,000 queries

Amazon Web Services has launched Web Search on Amazon Bedrock AgentCore as a generally available feature, giving AI agents access to a continuously updated proprietary web index without routing queries through third-party search engines.

What's new

The new managed connector gives Bedrock AgentCore agents real-time web access through an index Amazon maintains independently:

• Proprietary index: spans tens of billions of documents, refreshed continually so new content is reflected within minutes
• Knowledge graph integration: provides high-confidence responses for factual queries about entities and relationships
• Semantic snippet extraction: returns relevant passages optimized for model context windows rather than raw HTML
• Privacy design: queries remain entirely within AWS infrastructure and are not sent to any third-party search engine
• MCP-compatible: integrates with Model Context Protocol frameworks including Strands, LangChain, and LangGraph
• No infrastructure overhead: the managed connector requires no API provisioning or credential management

Pricing is $7 per 1,000 queries — less than one cent per question at typical agent call volumes. The feature is currently available in the us-east-1 (US East N. Virginia) region.

Context

Agent knowledge is frozen at training time, which has been a structural limitation for production deployments requiring current information. Existing options — passing third-party search API keys, building custom retrieval pipelines — require setup work and raise data-routing questions for regulated industries. AWS built its own index specifically to let enterprise customers add live web retrieval without sending query data outside their AWS environment.

Bedrock AgentCore, launched earlier this year, is Amazon's managed runtime for deploying AI agents at scale. Web Search is the latest in a series of built-in tools that reduce the infrastructure overhead of building production agents on AWS.

Why it matters

The privacy angle is the differentiator. AWS's pitch — "Customer queries don't get sent to a third-party search engine or leave AWS" — directly addresses enterprise procurement objections around data leakage that have slowed adoption of web-enabled agents in financial services, healthcare, and government. At $7 per thousand queries, the pricing is also low enough that retrieval-heavy agents can operate economically without complex caching.

The MCP compatibility is notable as well. By exposing web search as an MCP-compatible tool, AWS slots into the growing ecosystem of agent frameworks that speak the Model Context Protocol, letting teams swap or layer providers without rewriting agent logic. The launch accelerates a broader trend: managed web retrieval as a commodity, table-stakes capability for enterprise agent platforms rather than a differentiated feature.