AWS Reinvents AI Agents: Introducing Amazon Bedrock AgentCore
AWS continues to prove that the future of AI goes far beyond generating text. At the recent AWS Summit in New York, the company unveiled Amazon Bedrock AgentCore a new service providing infrastructure for deploying and managing AI agents. This solution is designed to empower AI agents to perform real tasks: connect to business tools, analyze data, process instructions, and make decisions not just respond to questions.
What is Amazon Bedrock AgentCore?
Amazon Bedrock AgentCore is an infrastructure framework for AI agents, created to support their deployment in production environments. Its primary goal is to simplify the process of building and operating AI agents that do more than just engage in conversations – they get things done.
Unlike some other AWS offerings, a key difference with AgentCore is that it does not provide language models (such as Claude, Mistral, Llama) directly, but integrates with them via Amazon Bedrock. AgentCore handles the entire operational environment around those models: agent sessions, memory, code interpretation, tool and internet access, authorization, and monitoring. It offers a set of modules that let you run AI agents in production without building custom architecture from scratch.
How does Amazon Bedrock AgentCore work?
The system is based on a cohesive architecture composed of seven modules, each responsible for a different aspect of the agent’s operation. The core component is the runtime – a scalable, serverless environment that launches an AI agent in an isolated and secure session. Each session can last up to eight hours, enabling complex interactions with persistent memory of previous steps and context.
AgentCore also includes built-in memory, storing data in two modes: short-term (for the current session) and persistent (for continuity across future user or process interactions). Importantly, companies don’t need to configure an external vector database or knowledge management system – memory is native to the platform.
The Gateway module acts as a translation layer, converting external resources like APIs, Lambda functions, or third-party tools (e.g., Jira, Slack, Salesforce) into the standard Model Context Protocol (MCP). This allows the agent to understand how to use a given tool, what data is accessible, and how to act within the correct context.
The Interpreter module enables agents to execute Python or JavaScript code in a sandboxed environment. That means the agent can calculate data, write scripts, or perform transformations independently – without human intervention.
Next is the Browser, a tool that gives the agent controlled internet access for real-time browsing and webpage analysis in a sandbox.
The Identity module ensures secure access and authorization, integrating with identity systems so the agent operates within defined policies – crucial for working with sensitive data, approval workflows, or regulated environments.
Finally, Observability ties it all together, offering a full set of monitoring, logging, and analytics tools to track agent behavior. From basic logs to detailed decision paths, telemetry, and performance indicators — teams get full visibility into the agent’s environment and can quickly respond to anomalies.
Who is it for and when is it available?
Amazon Bedrock AgentCore is intended for development teams, AI engineers, and organizations building custom agent-based solutions powered by language models. We believe it will be particularly useful for teams involved in customer support automation, application integration, data analysis, internal operations support, and R&D departments developing internal tools.
Currently, the service is available in Preview mode in selected AWS regions, including Northern Virginia (us-east-1), Oregon (us-west-2), Frankfurt (eu-central-1), and Sydney (ap-southeast-2).
Expanded availability is expected in the coming months of 2025, although no specific date has been announced yet. Interested users can begin testing now with AWS-provided documentation, SDKs, and sample deployments.
Follow us on LinkedIn to stay updated on the latest AWS services, real-world AI agent applications, and technologies that truly accelerate automation and cloud innovation.
