{"slug": "introducing-web-search-on-amazon-bedrock-agentcore", "title": "Introducing Web Search on Amazon Bedrock AgentCore", "summary": "Amazon Web Services announced the general availability of Web Search on Amazon Bedrock AgentCore, a fully managed, MCP-compatible web search capability that lets AI agents retrieve real-time information from a purpose-built web index without infrastructure overhead. The feature addresses the limitation of agents having knowledge frozen at training time by providing continuously updated search results, a knowledge graph for high-confidence facts, and semantic snippet extraction, all within AWS's privacy boundary.", "body_md": "[Artificial Intelligence](https://aws.amazon.com/blogs/machine-learning/)\n\n# Introducing Web Search on Amazon Bedrock AgentCore\n\nAI agents are changing how organizations find and act on information, but they share one structural limitation: their knowledge is frozen at training time. When you ask an agent that relies only on its training data about today’s stock price, a sports score, or a release that shipped an hour ago, it can’t respond.\n\n*Web Search* on [Amazon Bedrock AgentCore](https://aws.amazon.com/bedrock/agentcore/), now generally available, addresses that gap. This fully managed, Model Context Protocol (MCP)-compatible web search capability lets your agents get information from the web without infrastructure overhead. It’s available as a managed target or connector that you connect to your [AgentCore Gateway](https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/gateway.html). Agents discover it with a standard `tools/list`\n\ncall and invoke it like other MCP tools. There are no search APIs to provision, no outbound credentials to manage, and no result-parsing glue to maintain.\n\nBehind that single connector sits a purpose-built web index maintained by Amazon, spanning tens of billions of documents. Amazon refreshes the index continually, reflecting new content within minutes. The privacy model makes sure that queries don’t leave AWS. Retrieval can combine a knowledge graph with semantic snippet extraction tuned for model context.\n\nIn this post, we walk through what makes Web Search on Amazon Bedrock AgentCore different, why it matters, and how to wire it in with a few lines of code.\n\n**Figure 1:** Your application connects to the AgentCore Gateway (AWS Identity and Access Management (IAM) or JSON Web Token (JWT) inbound auth), which routes queries through a managed connector to the Web Search tool in the AWS service account. Query traffic stays within AWS.\n\nGrounding agents in the web is the fix for stale knowledge, but it’s also where many teams get stuck. Building it yourself means:\n\n- Procuring a third-party search API and managing keys, quotas, and rate limits.\n- Parsing inconsistent result formats across providers.\n- Reasoning about where customer queries travel and how that data might be retained or reused.\n- Building snippet extraction logic, so models get relevant passages, not raw HTML.\n- Maintaining freshness, coverage, and quality over time.\n\nEach of these is a project in itself. Web Search on Amazon Bedrock AgentCore addresses all of them.\n\n## A purpose-built web index\n\nMany “add web search to your agent” solutions are wrappers around a third-party search engine. Web Search on Amazon Bedrock AgentCore is backed by a web index that Amazon operates directly, spanning tens of billions of documents. That scale matters for coverage. For example, the long-tail question about a niche library or an obscure product spec can be answered more effectively when the index is broad rather than limited to the most popular pages.\n\n### Updated continually\n\nAmazon refreshes the index on an ongoing basis, reflecting new content within minutes. For agents that respond to questions about price movements or recently published announcements, that recency window is the difference between a grounded response and a confidently wrong one. When your agent searches for “what happened today,” the results reflect what actually happened today.\n\n### Knowledge graph for high-confidence facts\n\nWeb Search on Amazon Bedrock AgentCore includes a built-in knowledge graph that grounds entities and their relationships. For factual questions (like who holds a role or when something was founded), the knowledge graph provides high-confidence responses rather than leaving the model to infer them from extracted page text. This reduces the kind of subtle factual drift that creeps in when an agent stitches together a response from snippets alone.\n\n### Semantic snippet extraction optimized for model context\n\nRather than handing the model a raw HTML dump or a full page and hoping it finds the relevant part, the tool performs semantically relevant snippet extraction. It pulls the passages from each web page that bear on the query, then returns them in a form optimized for a model’s context window. The model sees the parts that matter, with fewer tokens spent on boilerplate and navigation chrome. This can help improve the precision of cited responses.\n\n## Private by design\n\nFor many enterprises, the question that stalls a web search rollout isn’t “does it work.” It’s “where do my users’ queries go, and what happens to them?” Web Search on Amazon Bedrock AgentCore is built so the answers to those questions are simple.\n\n### Queries don’t leave AWS\n\nWhen your agent issues a search, the query is served entirely within AWS infrastructure. Customer queries don’t get sent to a third-party search engine or leave AWS. The Gateway authenticates to the connector owned by AWS and routes the request internally, so the data path stays inside AWS end to end. For teams with data-residency or third-party egress concerns, this removes an entire category of review.\n\n## Walkthrough\n\nTo get started with the Web Search Tool, you create an AgentCore Gateway (if you don’t want to use an existing one), add the web search tool target, and invoke it from an agent using MCP.\n\n### Prerequisites\n\nTo follow along with the setup steps in this post, you need the following:\n\n- An AWS account with permissions to create IAM roles and Amazon Bedrock AgentCore resources.\n- The AWS Command Line Interface (AWS CLI) v2 installed and configured, or access to the AWS Management Console.\n- Python 3.10 or later (for the SDK and Strands examples).\n- The\n`boto3`\n\nSDK updated to the latest version. - An Amazon Bedrock AgentCore Gateway. You can add the Web Search Tool as a target to an existing Gateway, or create a new one. For instructions on creating a Gateway, see\n[Create an Amazon Bedrock AgentCore gateway](https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/gateway-create.html)in the Developer Guide.\n\nNote: Following these steps creates AWS resources that incur charges. The Amazon Bedrock AgentCore Gateway and Web Search invocations are billable. See the Pricing section that follows for details, and remember to clean up resources when finished to avoid ongoing charges.\n\n### Setup\n\nAdding web search to an agent comes down to attaching a Web Search Tool target to your Gateway using `connectorId: \"web-search\"`\n\n. The Gateway snapshots the tool schema, provisions the integration, and handles schema management, parameter governance, endpoint resolution, and service authentication for you.\n\nVerify that you added the target by calling `describe_gateway_target`\n\nor `list_gateway_targets`\n\nand confirming that Web Search-tool appears in the response.\n\n### The outbound role and permissions\n\nNotice the preceding `credentialProviderConfigurations`\n\n. This is the whole outbound-authorization story: instead of you provisioning API keys or managing search credentials, the Gateway authenticates to the Web Search backend using its own IAM service role.\n\nThat role needs a trust policy (so AgentCore can assume it, scoped to your account and Region) and a permissions policy with two actions:\n\nThe `InvokeWebSearch`\n\nresource ARN is owned by AWS (account = `aws`\n\n). Authorization is enforced per invocation against that ARN, so granting `bedrock-agentcore:InvokeWebSearch`\n\non it is what lets the Gateway call web search on your behalf.\n\nA couple of boundaries to keep clear:\n\n- This role is for outbound auth only (Gateway reaching the Web Search backend). Inbound auth (who can call your Gateway) is handled separately, typically with an OAuth or JWT authorizer such as Amazon Cognito.\n- The role doesn’t include\n`bedrock:InvokeModel`\n\n. Model access belongs to whatever identity runs your agent, not to the Gateway service role.\n\n### Invoking from MCP-compatible frameworks\n\nBecause Web Search is exposed over MCP, an MCP-compatible framework like Strands, LangChain, LangGraph, CrewAI, or your own can discover and invoke it. The agent calls `tools/list`\n\n, finds `WebSearchTool`\n\n, and uses it automatically whenever it needs current information:\n\nThe agent determines it needs fresh information, invokes `WebSearchTool`\n\nwith an appropriate query, and composes a grounded response with source citations. No tool-specific code on your side.\n\n### Response format\n\nResults come back in the standard MCP `tools/call`\n\nenvelope. The tool returns a single `content`\n\nblock of type `text`\n\nthat contains a serialized JSON document with the results. Parse that inner text and you get an `id`\n\nplus a `results`\n\narray of observations:\n\nEach web index observation (always returned) carries `title`\n\n, `url`\n\n, `publishedDate`\n\n, and `text`\n\n. Knowledge-graph observations (optional, for entity queries) have null `title`\n\nand `url`\n\nplus structured key/value facts in the `text`\n\nfield.\n\n## Where the Web Search Tool fits\n\nIf you need to ground an agent in your own enterprise data, Amazon Bedrock Knowledge Bases and Amazon Bedrock Managed Knowledge Bases are the right tools. They ingest, index, and retrieve over content you own. The Web Search Tool is the complement. It grounds agents in the public web, for questions whose responses live outside your organization and change by the minute. Many production agents use both: a knowledge base for “what do our documents say” and web search for “what’s true in the world right now.”\n\n## Pricing\n\nAt $7 per 1,000 queries, you can run a web-search agent for less than a cent per question with a pay-as-you-go model.\n\n## Clean up resources\n\nIf you created resources while following along, you can remove them to avoid ongoing charges:\n\n- Delete the Gateway target: call\n`delete_gateway_target`\n\nwith your`gatewayIdentifier`\n\nand`targetId`\n\n. - If the Gateway was created solely for this walkthrough, delete it with\n`delete_gateway`\n\n.\n\nThere is no persistent infrastructure on the AWS side beyond these resources. After they are removed, you stop incurring charges.\n\n## Conclusion\n\nThe Amazon Bedrock AgentCore Web Search Tool gives your agents current web knowledge through a single `connectorId`\n\n. There are no search APIs to provision and no result-parsing to maintain. Underneath that simplicity is a web index that AWS builds itself (tens of billions of documents, refreshed within minutes), a privacy model where queries don’t leave AWS, and retrieval that can combine a knowledge graph with semantic snippet extraction tuned for model context. The result is an agent that responds to timely questions accurately, cites its sources, and keeps your data where it belongs.\n\nBecause Amazon operates the full search stack, improvements to freshness, coverage, relevance, and snippet quality flow to your agents automatically through the same managed connector. No version upgrades or migrations are needed on your side.\n\nYou can access the Web Search Tool connector today in `us-east-1`\n\n(US East (N. Virginia)).\n\nTo get started, see the [Web Search Tool documentation](https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/gateway-target-connector-web-search-tool.html).", "url": "https://wpnews.pro/news/introducing-web-search-on-amazon-bedrock-agentcore", "canonical_source": "https://aws.amazon.com/blogs/machine-learning/introducing-web-search-on-amazon-bedrock-agentcore/", "published_at": "2026-06-19 14:15:24+00:00", "updated_at": "2026-06-19 14:36:41.043546+00:00", "lang": "en", "topics": ["ai-agents", "ai-infrastructure", "ai-tools", "large-language-models", "artificial-intelligence"], "entities": ["Amazon Web Services", "Amazon Bedrock AgentCore", "Model Context Protocol", "AgentCore Gateway", "AWS Identity and Access Management", "JSON Web Token"], "alternates": {"html": "https://wpnews.pro/news/introducing-web-search-on-amazon-bedrock-agentcore", "markdown": "https://wpnews.pro/news/introducing-web-search-on-amazon-bedrock-agentcore.md", "text": "https://wpnews.pro/news/introducing-web-search-on-amazon-bedrock-agentcore.txt", "jsonld": "https://wpnews.pro/news/introducing-web-search-on-amazon-bedrock-agentcore.jsonld"}}