AWS adds Web Search to Bedrock for AI models to tap live information

AWS announced general availability of Web Search on Amazon Bedrock at AWS New York Summit 2026, extending a feature the company had already released on AgentCore. Web Search is a server-side built-in tool designed to ground model responses in current web knowledge, eliminating the need for third-party vendors, external APIs, or additional security reviews.

The tool works through a multi-source grounding approach. It is backed by a web index that Amazon operates, spanning billions of documents and refreshed continually, combined with a built-in knowledge graph that anchors entities and connections between them. When a question is factual—such as who wrote a particular book or what year an event took place—Web Search uses the knowledge graph to answer with strong confidence rather than leaving the model to infer from extracted page text. This approach reduces small factual inaccuracies that slip in when an agent assembles answers from fragments. The tool also performs semantic snippet extraction, pulling only the passages from each web page that bear on the query and returning them in a form optimized for the model's context window, which reduces token overhead and latency.

Enabling Web Search requires three configuration steps. First, developers configure AWS credentials and permissions: the tool uses existing AWS IAM credentials (no separate API keys), and the calling identity must have inference permissions on Amazon Bedrock and Web Search tool permissions (at minimum, bedrock-websearch:InvokeSearch; optionally bedrock-websearch:InvokeFetch to let the model read full page content). Second, they point the OpenAI client at the bedrock-mantle endpoint and add the Web Search tool to the request with a single parameter: `tools=[{"type": "web_search", "external_web_access": False}]`. The external_web_access field selects whether Web Search retrieves from Amazon's pre-indexed web corpus or live content fetched directly from the web; currently only indexed-web retrieval is served, though live-web retrieval will be enabled in a future update. Third, the API returns a grounded response with structured citation annotations, including URL and page title for each referenced source.

When Web Search is enabled in an API call, Bedrock handles the entire search lifecycle server-side. The model identifies that a query requires up-to-date web knowledge, Bedrock formulates a search query, retrieves relevant content from Amazon's web index and knowledge graph, and injects results (including snippets, source URLs, and titles) into the model's context window. The model reasons over the retrieved content and generates a grounded response with source citations. The API returns the final response in a single round-trip—there is no client-side tool-use loop to build, no external API responses to parse, and no retries or rate limits to manage.

Web Search is integrated with AWS CloudTrail out of the box, recording every call to bedrock-websearch:InvokeSearch and bedrock-websearch:InvokeFetch as a management event, capturing the calling identity, timestamp, action, and account and Region context. Access-denied outcomes are always logged, with the specific condition key that caused the denial included, making IAM misconfigurations easy to diagnose. By design, CloudTrail does not record query text, the URLs returned by search, or raw page content retrieved by fetch—query text is treated the same way as an inference prompt and is never exposed in trail events. Combined with in-Region processing and zero data egress, this provides a full audit trail without exposing what end users searched for.

Web Search on Bedrock is generally available in US regions, with in-region query handling in us-east-1, us-east-2, and us-west-2. Pricing details are available on the Amazon Bedrock pricing page.