Agents can pull fresh web info now! Web Search lands on Bedrock AgentCore!

Hey everyone, it's Shii! Today I found some news that's going to make AI agents a lot smarter, and I got so excited. There's a new way for agents to look up current information on the web themselves and answer with sources attached. Let me tell you all about it!

AWS Blog aws.amazon.com

What was announced?

AWS's official blog (AWS Blog) announced that Web Search has arrived on Amazon Bedrock AgentCore. It's a fully managed tool for AI agents, and it's now generally available (GA).

When an agent sends a natural-language query, Web Search returns relevant snippets, source URLs, titles, and publication dates. The important part is that the search stays inside AWS: there's zero data egress from your secured AWS environment. You can ground your agent's answers in current information without sending queries to an external search provider.

Why it matters

AI agents have a weakness: they don't know anything newer than their training cutoff. So whenever you need "today's information" — fresh news, prices, updated docs — you really want to lean on web search.

But wiring up your own web search inside an agent means integrating with an external API and managing the infrastructure, which is a lot of work. And for companies, there's the security worry of not wanting to send search queries outside. This Web Search tool bridges those two needs: wanting current info, and not wanting data to leave your environment.

What changes

The web search plumbing you used to build by hand is now available as a managed tool on the AgentCore side. Instead of babysitting infrastructure, developers can focus on building the agent itself.

Because answers come with source URLs, titles, and publication dates, it's clear where a piece of information came from, which makes it easier to check how trustworthy an agent's answer is. The article shares voices from customers like Benchling, an R&D platform, and Gen Digital, a cybersecurity company. Gen Digital especially values that AWS uses its own search index and keeps queries inside its trusted AWS environment.

Let's dive deeper!

Let's look a bit more closely at how it works!

Web Search is built on Amazon's search infrastructure and talks to agents using the Model Context Protocol (MCP). Instead of just querying a web index, it uses "multi-source grounding" that also combines structured data from the Amazon Knowledge Graph to return verified facts. This is the same technology used in Alexa+, Amazon Quick, and Kiro.

Getting started is simple: create a Bedrock AgentCore Gateway in the console, choose "MCP target" as the protocol and "Connectors" as the target type, and pick the preconfigured "Web Search tool." You can add it to a new Gateway or an existing one. There are plenty of ways to call it, too — the API, the CLI, MCP Inspector, Python code with the MCP SDK, or the Strands MCP Client.

As for pricing, it's usage-based at $7 per 1,000 search queries. There are no upfront commitments, and new AWS customers get up to $200 in Free Tier credits.

For availability, it starts in US East (N. Virginia). You can check the roadmap for other Regions in AWS Capabilities by Region.

Wrap-up

Let me recap today's highlights!

• Web Search, a fully managed tool for agents, is now generally available on Amazon Bedrock AgentCore
• Search stays inside AWS, with zero data egress from your secured environment
• It returns snippets, URLs, titles, and publication dates, and combines the Amazon Knowledge Graph for grounding
• You can add it just by choosing the Web Search tool via "MCP target" and "Connectors" on a Gateway
• Pricing is $7 per 1,000 searches, starting in US East (N. Virginia)

If you want to build agents on AWS that are strong on current information — especially if you're security-conscious and don't want queries leaving your environment — this is definitely one to check out!