We’re excited to announce that Web Search is now generally available on Amazon Bedrock AgentCore. This fully managed tool allows AI agents to base their answers on up-to-date, cited web information, all without any data leaving your secure AWS environment.
Web Search operates through a built-in connector within the Bedrock AgentCore Gateway, utilizing the Model Context Protocol (MCP). When your agent submits a query in natural language, Web Search provides the most relevant content snippets, along with source URLs, titles, and publication dates. The AI model then uses this information to generate accurate, well-supported responses.
This feature is powered by Amazon’s robust search infrastructure, drawing from years of expertise in supporting intelligent search across services like Alexa+, Amazon Quick, and Kiro. It employs a multi-source approach that blends Amazon’s web index with structured data from a knowledge graph. In addition to standard web results, agents gain access to the Amazon Knowledge Graph, which contains verified facts, enabling them to deliver more precise and reliable answers than conventional web searches.
With this release, you can concentrate on developing your agents rather than handling the complexities of integrating and managing web search capabilities within Bedrock AgentCore. Your AI agent can now interpret user questions, fetch the latest information, and take informed actions based on current events that extend beyond the model’s original training data. Furthermore, you can comply with enterprise governance requirements by ensuring that user prompts and search queries remain within the AWS ecosystem, avoiding the need to send data to third-party search APIs.
Getting Started with Web Search on Bedrock AgentCore
To begin, set up a Bedrock AgentCore Gateway and add the Web Search tool as a target via the Bedrock AgentCore console. Once your Gateway URL is ready, you can interact with it using API calls, the Command Line Interface (CLI), or the MCP Inspector.
When configuring your Gateway, select MCP target as the protocol and Connectors as the target type. Then, choose the Web Search tool as a preconfigured option to fetch relevant web results, including links, snippets, and metadata.
After your gateway is created, you’ll find the Web Search tool listed on its details page. You can also add this tool to an existing gateway at any time.
To use the Web Search tool, refer to the sample code provided in the View invocation code section. You can integrate it using Python with API requests, the MCP Python SDK, the Strands MCP Client, or the MCP Inspector.
For instance, you can use the MCP Inspector—a developer tool for testing and debugging MCP servers. By connecting to the MCP server via the Gateway resource URL, you’ll see the Web Search tool available for each connector target. Simply enter your search query and click Run Tool to retrieve results.
For detailed instructions on using Web Search with Bedrock AgentCore, consult the official Bedrock AgentCore Gateway documentation.
What Customers Are Saying
Several customers had early access to this feature. Here’s their feedback:
Benchling supports scientists in accelerating research and development by centralizing data, enabling collaboration, and providing insights. Nicholas Larus-Stone, Head of AI Agents at Benchling, noted, “With Benchling AI, scientists can inquire about their current research targets and receive answers grounded in both their internal Benchling data and published literature. This leads to more comprehensive science and better hypothesis generation. Thanks to the Web Search tool on Amazon Bedrock AgentCore, our customers can securely integrate high-quality published data into their workflows while maintaining strict data governance.”
Gen Digital is a leader in consumer and small business cybersecurity, offering solutions like antivirus, identity protection, VPNs, and cloud backup. Iskander Sanchez-Rola, Senior Director of AI & Innovation at Gen Digital, shared, “The Web Search tool on Amazon Bedrock AgentCore empowers Norton Revamp to help professionals enhance their online reputation with timely, relevant content ideas based on real-world developments. We particularly appreciate that AWS uses its own search index and keeps all queries within our trusted AWS environment.”
Discover more customer success stories on the Amazon Bedrock Customers page.
Availability and Pricing
Web Search on Amazon Bedrock AgentCore is now generally available in the US East (N. Virginia) Region. Check the AWS Capabilities by Region page for details on regional availability and future updates.
Getting started is straightforward with no upfront costs. Pricing is usage-based, calculated per search query submitted by your agents. The rate is $7 per 1,000 queries. New AWS customers can also take advantage of up to $200 in Free Tier credits. For full pricing details, visit the Amazon Bedrock AgentCore pricing page.
Try it out in the Amazon Bedrock AgentCore console and share your feedback via AWS re:Post for Amazon Bedrock AgentCore or through your regular AWS Support channels.
— Channy
Updated on June 18, 2026 — Added a clear pricing statement for Web Search in Bedrock AgentCore.
