SEATTLE, WA – Amazon Web Services (AWS) today announced the general availability of Web Search on Amazon Bedrock AgentCore, a significant advancement designed to empower AI agents with real-time, cited web knowledge without compromising data security or necessitating data egress from customer environments. This fully managed tool represents a crucial step in addressing the inherent limitations of large language models (LLMs), particularly their static training data and propensity for generating inaccurate or "hallucinated" responses when faced with queries requiring current information. The integration enables developers to build more reliable, accurate, and contextually aware AI agents, streamlining workflows and bolstering decision-making across various enterprise applications.
The new Web Search capability is engineered to seamlessly integrate with agents built on Bedrock AgentCore, AWS’s comprehensive service for creating and managing AI agents that can perform complex tasks, interact with enterprise systems, and leverage multiple tools. By utilizing a built-in connector target on the Bedrock AgentCore Gateway via the Model Context Protocol (MCP), agents can now dispatch natural-language queries to Amazon’s robust search infrastructure. This infrastructure, refined over years of powering sophisticated agentic search experiences across prominent Amazon products like Alexa+, Amazon Quick, and Kiro, is designed to retrieve the most relevant snippets, source URLs, titles, and publication dates. The AI model then reasons over this freshly retrieved information to produce a grounded and verifiable response, directly mitigating the risk of outdated or fabricated outputs.
Technical Underpinnings and Multi-Source Grounding
At the core of this innovation lies Amazon’s proprietary search infrastructure, a sophisticated system built upon decades of experience in information retrieval. Unlike generic web search APIs, this integrated solution leverages a multi-source grounding approach. It combines a vast, continuously updated web index with structured knowledge graph data, specifically Amazon Knowledge Graph. This dual approach is critical: while the web index provides broad, real-time coverage of public information, the Amazon Knowledge Graph contributes verified facts and relationships, significantly enhancing the accuracy and relevance of search results. This combination allows agents to access a richer, more reliable dataset, often yielding superior results compared to relying solely on traditional web search methods. The Model Context Protocol (MCP) facilitates this interaction, serving as the standardized communication framework that allows agents to send natural language queries and receive structured search results efficiently and securely within the AWS ecosystem. This ensures that the data flow remains entirely within the customer’s secured AWS environment, a critical feature for enterprises with stringent data governance and compliance requirements.
Addressing the Challenges of Generative AI
The introduction of Web Search on Bedrock AgentCore directly addresses some of the most pressing challenges in the burgeoning field of generative AI. Large language models, while powerful in their ability to understand and generate human-like text, are fundamentally limited by the data they were trained on. This static knowledge base means they can quickly become outdated, struggle with real-time events, and are prone to "hallucinations"—generating confident but factually incorrect information. The "grounding" of LLM responses in verifiable, current external data has emerged as a paramount concern for enterprise adoption.
Prior to this release, developers building AI agents on Bedrock AgentCore often had to construct intricate architectures involving external search APIs or custom retrieval-augmented generation (RAG) pipelines. These solutions, while functional, introduced complexities related to integration, infrastructure management, and crucially, data security. Sending user prompts and retrieval queries to third-party search providers outside of the AWS trusted environment posed significant data egress risks and compliance hurdles for many organizations, particularly those in highly regulated industries. AWS’s integrated Web Search tool removes this burden, providing a pre-configured, secure, and managed solution that simplifies the development process and ensures data remains within the AWS perimeter. This strategic move by AWS underscores its commitment to providing a holistic and enterprise-grade platform for building responsible and reliable AI applications.
Amazon’s Legacy in Agentic Search Experiences
The capabilities embedded in Web Search on Bedrock AgentCore are not nascent but are the culmination of Amazon’s extensive experience in developing and deploying agentic search systems. The technologies powering this new feature are informed by years of research and operational insights gained from products like Alexa+, Amazon Quick, and Kiro.
- Alexa+: As a sophisticated AI assistant, Alexa has long relied on robust search and information retrieval mechanisms to answer user queries, control smart home devices, and provide real-time updates. The ability to understand natural language and fetch relevant information instantly, often from dynamic web sources, is a core component of Alexa’s functionality.
- Amazon Quick: This enterprise search solution helps organizations find information across their various data sources. Its effectiveness hinges on efficient indexing, sophisticated query processing, and the ability to surface relevant results quickly from diverse datasets.
- Kiro: An internal Amazon service focused on knowledge management and information retrieval, Kiro further exemplifies Amazon’s prowess in structuring and accessing vast amounts of data for internal operational efficiency and decision support.
This rich heritage ensures that the Web Search component of Bedrock AgentCore is not merely a basic search function but a highly optimized, scalable, and resilient system designed for complex, real-world agentic interactions. The "years of experience" translate into a search engine that is adept at disambiguation, relevance ranking, and handling varied query types, all critical for effective AI agent performance.
Operational Benefits and Enterprise Governance
For enterprises, the benefits extend beyond just technical capability. The Web Search on Bedrock AgentCore offers significant operational advantages and addresses critical governance concerns:
- Simplified Development: Developers can now focus on the core logic and unique value proposition of their AI agents rather than spending resources on integrating and managing external web search infrastructure. The built-in connector on AgentCore Gateway means web search can be enabled with a few clicks, drastically reducing development cycles.
- Real-time Relevance: Agents can now access the latest information, ensuring their responses are grounded in current developments rather than being limited to the model’s static training data. This is crucial for applications requiring up-to-the-minute data, such as financial analysis, news summarization, customer support, or scientific research.
- Enhanced Accuracy and Reliability: By reasoning over verifiable web snippets and knowledge graph data, agents can produce more accurate and trustworthy responses, reducing the incidence of hallucinations and improving user confidence.
- Robust Data Governance and Security: This is perhaps one of the most compelling features for enterprise adoption. By keeping all user prompts and retrieval queries within the customer’s secured AWS environment, the solution eliminates data egress risks associated with third-party APIs. This inherent security and compliance architecture is vital for industries with strict regulatory requirements, such as healthcare, finance, and government. It provides a secure sandbox for AI operations, protecting sensitive information and intellectual property.
- Scalability and Performance: Leveraging Amazon’s proven search infrastructure, the Web Search tool is designed to scale efficiently to meet the demands of enterprise-level AI deployments, ensuring consistent performance even under heavy loads.
Implementation and Developer Experience
Activating Web Search on Bedrock AgentCore is designed to be straightforward for developers. The process begins in the Bedrock AgentCore console, where users create a Bedrock AgentCore Gateway. During this creation, developers select "MCP target" as the protocol and "Connectors" as the target type. From the preconfigured options, the "Web Search tool" can be chosen to retrieve relevant web search results, including links, snippets, and metadata.
Once the Gateway is established, the Web Search tool target appears on its detail page, allowing for easy management and modification. Developers can also add the Web Search tool to existing gateways. To interact with the tool, sample invocation code is provided, supporting various methods including Python SDK, API requests, Strands MCP Client, and the MCP Inspector. The MCP Inspector, an interactive developer tool, allows for real-time testing and debugging of MCP servers. By connecting to the MCP server via the Gateway resource URL, developers can input web search queries and instantly view the results, facilitating rapid iteration and validation of agent behavior. This developer-centric approach aims to minimize friction and accelerate the deployment of intelligent agents. Comprehensive documentation is available on the Bedrock AgentCore Gateway documentation pages, offering detailed guidance on configuration and usage.
Early Adopter Insights and Real-World Applications
Early access partners have already begun to leverage the Web Search capabilities, demonstrating its immediate impact across diverse industries.
Benchling, a leading platform for scientific R&D, found the integration particularly valuable. Nicholas Larus-Stone, Head of AI Agents at Benchling, articulated the transformative potential: "Scientists using Benchling AI can now ask about a target they’re actively working on and get answers grounded in both their institutional data in Benchling and published literature. The result is more complete science, and hypothesis generation done right. Because we’re using the Web Search tool on Amazon Bedrock AgentCore, customers have a secure, governed environment to bring that high quality published data into their workflows without compromising how they manage their data." This highlights the critical need for scientific accuracy and the ability to synthesize proprietary institutional knowledge with the latest published research, all within a secure, compliant framework. For scientific discovery, where precision and verifiable data are paramount, Web Search provides a crucial bridge between internal datasets and the vast external body of scientific knowledge.
Gen Digital, a global leader in cyber safety, offering products like Norton, also lauded the new feature. Iskander Sanchez-Rola, Senior Director of AI & Innovation, Gen Digital, commented, "With the Web Search tool on Amazon Bedrock AgentCore, Norton Revamp helps professionals build their online reputation with current, grounded content ideas shaped by what’s actually happening in the world today. What we value most is that AWS uses its own search index and keep queries within our trusted AWS environment." This testimonial underscores the value for dynamic content generation and marketing, where staying current with global events and trends is essential for relevance. Furthermore, Gen Digital’s emphasis on AWS utilizing its own search index and maintaining queries within their trusted environment reinforces the security and privacy advantages for enterprises handling sensitive user data and content strategies.
These early examples illustrate how the Web Search tool is not merely a technical add-on but a strategic enabler for enterprises seeking to deploy AI agents that are both powerful and trustworthy. From accelerating scientific breakthroughs to crafting timely digital content, the ability to securely ground AI responses in real-time web knowledge is proving invaluable.
Market Impact and Future Outlook
The general availability of Web Search on Amazon Bedrock AgentCore is poised to have a significant impact on the broader AI ecosystem. It strengthens AWS’s position as a leading provider of comprehensive generative AI services, offering a robust platform that addresses critical enterprise needs for accuracy, security, and ease of development. This move intensifies competition among cloud providers, pushing the boundaries of what integrated AI platforms can offer.
For developers and organizations, it democratizes access to advanced grounding capabilities, reducing the barrier to entry for building sophisticated AI agents. The focus shifts from infrastructure management and data security concerns to innovative application development and strategic problem-solving. As AI agents become more prevalent in automating complex tasks, from customer service and data analysis to content creation and research, the ability to access and synthesize real-time, verifiable information will be non-negotiable.
Looking ahead, this release sets the stage for further enhancements in agentic AI. AWS is likely to continue expanding the range of tools and connectors available through Bedrock AgentCore, enabling agents to interact with an even wider array of enterprise systems, specialized databases, and external services. The emphasis on secure, in-environment processing could also inspire new standards for enterprise AI governance and trust. This is not just an incremental update but a foundational component for the next generation of intelligent, autonomous AI systems.
Availability and Pricing
Web Search on Amazon Bedrock AgentCore is generally available today in the US East (N. Virginia) Region, with plans for expanded regional availability to be outlined in future roadmap updates. Customers can monitor the AWS Capabilities by Region page for the latest information.
Crucially, customers can get started with Web Search on Bedrock AgentCore at no additional cost for the tool itself. Pricing is structured around the data transfer charges associated with using the Gateway. New AWS customers are also eligible for up to $200 in Free Tier credits, providing an accessible entry point for experimentation and deployment. Detailed pricing information is available on the Amazon Bedrock AgentCore pricing page. AWS encourages developers and enterprises to explore the capabilities through the Amazon Bedrock AgentCore console and provide feedback via AWS re:Post for Amazon Bedrock AgentCore or their usual AWS Support contacts. This release marks a pivotal moment in making AI agents more intelligent, reliable, and secure for enterprise applications worldwide.
