Amazon Web Services (AWS) has announced the general availability of Amazon Bedrock AgentCore, a platform designed to help developers build, deploy, and operate enterprise-grade artificial intelligence agents. The service aims to provide the necessary infrastructure for security, scalability, and reliability, moving AI agents from experimental phases to full-scale production environments.
AgentCore is now accessible in nine AWS Regions and offers a suite of managed services that support the entire agent development lifecycle. The platform is designed to be flexible, allowing developers to use their preferred AI models and frameworks while leveraging AWS's operational foundation.
Key Takeaways
- General Availability: Amazon Bedrock AgentCore is now fully available for production use across nine global AWS regions.
- Comprehensive Platform: AgentCore provides a full suite of managed services for building, deploying, and operating AI agents, including tools for code execution, web browsing, and API integration.
- Enterprise-Grade Features: The platform emphasizes security through isolated environments, scalability from zero to thousands of sessions, and observability with detailed audit trails.
- Flexibility: Developers can use various AI models (from Bedrock, OpenAI, Gemini) and frameworks (like CrewAI, LangGraph, LlamaIndex) with AgentCore's services.
- Industry Adoption: Companies like Sony, Ericsson, and Cohere Health are already using AgentCore to automate complex workflows and improve operational efficiency.
A Platform for Production-Ready AI Agents
Developing AI agents for business use often involves significant challenges beyond initial prototyping. Companies require a stable and secure foundation to manage these agents, which can perform actions and access sensitive data. Amazon's AgentCore is positioned as a solution to this problem, offering a managed platform that handles the underlying operational complexities.
The service is structured to support developers at every stage, from building the agent's core logic to deploying it at scale and monitoring its performance. This integrated approach is intended to accelerate the transition of AI agents into practical business applications.
Core Capabilities of AgentCore
AgentCore is composed of several distinct but interoperable services. Organizations can select the components they need or use the entire suite for a complete solution. This modularity allows teams to integrate AgentCore with their existing tools and workflows.
The platform is designed to be agnostic to specific models or frameworks. It supports popular choices like CrewAI, LangGraph, and the OpenAI Agents SDK, as well as models available on Amazon Bedrock and external models like OpenAI's GPT series and Google's Gemini.
What Are AI Agents?
AI agents are autonomous programs that can understand goals, make decisions, and perform actions to achieve those goals. Unlike simple chatbots, agents can use tools, access data, and execute multi-step tasks, such as booking travel, managing customer support tickets, or automating software testing.
Essential Tools and Infrastructure
To perform useful tasks, AI agents need access to various tools and systems. AgentCore provides several foundational services to enable these capabilities securely and efficiently.
Enabling Agent Actions
A key part of the platform is a set of pre-built tools for common agent tasks:
- AgentCore Code Interpreter: This service allows an agent to generate and run code in a secure, isolated sandbox environment.
- AgentCore Browser: Enables agents to interact with websites and web applications to gather information or complete tasks.
- AgentCore Gateway: Acts as a central point for connecting agents to a company's internal APIs and external third-party services like Jira or Zendesk.
- AgentCore Identity: Manages authentication and authorization, ensuring agents have the correct permissions to access systems on behalf of a user, utilizing standards like OAuth.
Scaling and Session Management
The AgentCore Runtime is built to handle unpredictable workloads. It can automatically scale from zero to thousands of concurrent sessions and supports long-running tasks for up to eight hours, a feature critical for complex workflows.
Memory and Observability
For agents to be effective, they must remember past interactions and context. AgentCore Memory provides a managed service for this, allowing agents to maintain conversational history and user preferences without developers needing to build and manage complex memory systems.
Because agents can act in non-deterministic ways, monitoring their behavior is crucial. AgentCore Observability offers real-time dashboards and detailed audit trails to track every action an agent takes. This helps with debugging, performance optimization, and ensuring compliance. It is compatible with standard monitoring tools like Datadog and LangSmith through its support for OpenTelemetry (OTEL).
Security and Enterprise Readiness
Security is a primary focus of the AgentCore platform, as agents often need to access sensitive corporate data and systems. The architecture incorporates several layers of protection to ensure secure operations.
"Agents that companies are willing to bet their business on need an enterprise-grade operational foundation—one that is secure, reliable, scalable, and purpose-built for the non-deterministic nature of agents." - Swami Sivasubramanian, Vice President for Agentic AI at AWS.
Each agent session runs in its own isolated computing environment using microVM technology. This prevents potential data leaks or interference between different agent tasks. Furthermore, AgentCore supports deployment within a Virtual Private Cloud (VPC) and uses AWS PrivateLink, ensuring that network traffic remains private and does not traverse the public internet.
Real-World Applications and Customer Impact
Several major companies have already started using AgentCore to build and deploy AI agents for various business needs, demonstrating the platform's utility across different industries.
Transforming Complex Industries
In the highly regulated healthcare sector, Cohere Health built an AI copilot named Cohere Review Resolve™ using AgentCore. This tool assists with medical necessity reviews by analyzing clinical records to find evidence supporting treatment requests. The company expects the solution to reduce review times by 30-40% and improve the accuracy of clinical decisions by approximately 30%.
Telecommunications leader Ericsson is using AgentCore to manage its vast and complex 3G/4G/5G/6G systems, which consist of millions of lines of code. Dag Lindbo, Head of AI at Ericsson, noted that AgentCore helps them fuse data and information to create capable AI agents that can scale across a workforce of tens of thousands.
Enhancing Entertainment and Manufacturing
Sony Group has adopted AgentCore to build a company-wide Agentic AI Platform. Masahiro Oba, a Senior General Manager at Sony, stated that the platform provides enterprise-level security, observability, and scalability, which are critical for accelerating the company's AI transformation securely.
Within Amazon itself, the Devices Operations & Supply Chain team is pioneering an "agentic manufacturing" approach. AI agents collaborate to automate processes, such as generating quality control test procedures from product specifications. This has reduced the time to fine-tune an object detection model from several days to less than an hour.
Global Availability and Getting Started
Amazon Bedrock AgentCore is now generally available in nine AWS Regions: US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (Dublin), Europe (Frankfurt), Asia Pacific (Mumbai), Asia Pacific (Singapore), Asia Pacific (Sydney), and Asia Pacific (Tokyo).
Developers can begin building with the platform immediately. The AgentCore SDK has reportedly been downloaded over one million times, indicating strong interest from the developer community. AWS is also facilitating adoption through the AWS Marketplace, where partners can offer pre-built agents and tools designed to work with AgentCore.