AWS Sells Direct Web Search at $7 per 1,000 Queries
AWS launched a standalone Web Search API for Bedrock AgentCore on November 11, 2026, billing $7 per 1,000 queries and removing the gateway requirement for custom agents and pipelines.
By Olivia Hart
4 min read
What's News
- AWS announced the standalone Web Search API for Amazon Bedrock AgentCore on November 11, 2026.
- Searches are billed at $7 per 1,000 queries, with a default quota of 10 requests per second per account per Region.
- Queries run against an Amazon-operated index on the same infrastructure as Alexa+, Kiro and Quick.
- Launch availability covers six regions: N. Virginia, Ohio, Oregon, Ireland, Tokyo and Sydney.
- CloudTrail logs InvokeWebSearch only as non-default data events, recording tool and maxResults but not search text.
Amazon Web Services now charges $7 for every 1,000 web searches run through a standalone API, a price the company set when it launched direct web search for Amazon Bedrock AgentCore on November 11, 2026. One signed call returns ranked web results with no gateway to build and no third-party API key. That works out to seven-tenths of a cent per query — cheap for a dashboard refresh, material for a pipeline running millions of lookups a day.
The new InvokeWebSearch operation replaces what used to be a three-part setup. Previously, Web Search on AgentCore required the AgentCore Gateway, a web search target, and a Model Context Protocol (MCP) client. Now a single request through AWS SDKs or the AWS CLI, authenticated with standard IAM credentials and Signature Version 4, returns text extracts plus page titles, URLs and publication times where available.
Who is the direct API for?
AWS targets custom agent loops and frameworks such as LangChain or LlamaIndex, plus non-agent workloads: dashboards, fact-checking pipelines, content enrichment and monitoring systems. Developers also take ownership of retries, caching and error handling — functions the gateway previously handled for them.
The service searches an Amazon-operated web index. AWS says queries are served inside AWS rather than forwarded to a third-party search engine, and that the operation runs on the same search infrastructure as Alexa+, Kiro, Quick and other Amazon products. The company has not disclosed the index's size, freshness or coverage, so buyers pricing this against external search APIs are comparing on cost alone.
The technical limits are explicit:
- Queries up to 200 characters
- Callers request 1 to 25 results; omitting maxResults returns up to 25
- Domain include and exclude lists of up to 100 entries each, with a root domain matching its subdomains
- Publication filters accept UTC timestamps in RFC 3339 format, with inclusive publishedAfter and publishedBefore bounds
Python callers need Boto3 1.43.127 or later, with support also in langchain-aws 1.9.3 and AgentCore Python SDK 1.26.0. The AgentCore console lets users test queries with result-count, date and domain settings, inspect result metadata and call latency, then generate an integration snippet or create a gateway target. In short: AWS is pushing experimentation before commitment.
What does the API change for governance?
Callers need the bedrock-agentcore:InvokeWebSearch permission on the AWS-managed web search tool — the same action and resource a gateway's service role uses. Service control policies can apply guardrails across accounts, letting organizations govern direct search alongside existing AWS permissions.
For private-subnet workloads, AWS PrivateLink routes calls through an interface endpoint in a VPC, keeping traffic on the AWS network without an internet gateway, NAT gateway or proxy. Endpoint policies, security groups, subnet placement and an aws:SourceVpce IAM condition can restrict which principals and network paths invoke search. The caveat is administrative weight: every one of these controls is opt-in.
Audit coverage has a gap worth noting. CloudTrail records InvokeWebSearch requests as data events, but these are not logged by default. A trail or event data store needs an event selector for the AWS::BedrockAgentCore::WebSearch resource type. Even then, the events identify only the web search tool and maxResults — not the search text or filters. Compliance teams get volume, not content. The acceptable-use terms also require customers to retain and display the source citations and links supplied with results, a legal obligation that sits with the buyer.
Where is it available, and when should you stay on the gateway?
The direct API meters to the same usage type as the gateway connector but incurs no gateway charges. The default quota is 10 requests per second per account in each Region, and customers can request an increase. At launch, the service covers US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (Ireland), Asia Pacific (Tokyo) and Asia Pacific (Sydney) — six regions, with no announced timeline for expansion.
AWS still recommends AgentCore Gateway for most agentic workloads, particularly deployments that combine MCP-compatible tools behind one endpoint, use optional MCP sessions, or apply one administrator-controlled source policy across teams. Both routes search the same index, but the gateway can authorize callers with JSON Web Tokens, while the standalone API uses IAM only. Gateway administrators can enforce target-level domain lists hidden from calling agents; direct API requests carry their own filters, visible to whoever writes the code.
The dividing line is operational control. Direct callers own request handling and filtering; shared agent environments keep the gateway's centralized tool, authentication and source-policy controls. The launch makes Amazon's web index accessible without adopting MCP or the gateway stack — the trade is convenience against centralized oversight.
Source: unite.ai
More from Olivia Hart
Show full bio
Staff writer covering people, launches and day-to-day practice in entrepreneurship, startups and venture funding at Business Bearings.
2,008 articles