Amazon Bedrock AgentCore
Amazon Bedrock AgentCore is an agentic platform to build, deploy, and operate highly capable agents securely at scale using any framework, model, or protocol. Using AgentCore, you can do the following, all without any infrastructure management:
-
Build agents faster.
-
Enable agents to take actions across tools and data.
-
Run agents securely with low-latency and extended runtimes.
-
Monitor agents in production.
AgentCore eliminates the undifferentiated heavy lifting of building specialized agent infrastructure, allowing you to accelerate your agents to production. Its services can be used together or independently and are compatible with any framework, including CrewAI, LangGraph, LlamaIndex, and Strands Agents. AgentCore is also compatible with any foundation model that’s available in or outside of Amazon Bedrock, providing ultimate flexibility.
AgentCore is composed of several key services:
-
Amazon Bedrock AgentCore Runtime – Provides a secure, serverless, scalable environment to host and run your agents, without needing to manage any infrastructure required for deploying and running AI agents or tools.
-
Amazon Bedrock AgentCore Memory – Offers a managed memory system, enabling agents to retain context from interactions for more personalized and coherent conversations by maintaining both immediate and long-term knowledge.
-
Amazon Bedrock AgentCore Gateway – Simplifies the process of creating, securing, and finding the right tools for agents. With AgentCore Gateway, developers can convert APIs, Lambda functions, and existing services into Model Context Protocol (MCP)-compatible tools and make them available to agents.
-
Amazon Bedrock AgentCore Identity – Provides a secure, scalable agent identity and access management service that accelerates AI agent development. With AgentCore Identity, you can assign unique, verifiable identities to agents, enabling fine-grained access control and secure agent-powered interactions with enterprise systems.
-
Amazon Bedrock AgentCore built-in tools – Enables you to use built-in tools to enhance your development and testing workflow. Use these tools to interact with your application effectively, enabling AI agents to write and execute code securely in sandbox environments. Use the browser tool to enable AI agents to interact with websites at scale.
-
Amazon Bedrock AgentCore Observability – Delivers logging and monitoring capabilities, giving you real-time visibility into your agent's performance and behavior to facilitate debugging and optimization.
Key features of AgentCore
AgentCore includes the following key features:
-
Fully managed and extensible – AgentCore is a fully managed service, which means that AWS handles the underlying infrastructure and maintenance. It is also extensible, which allows you to customize and enhance the functionality of your agents. For more information, see Get started with AgentCore Runtime in the AgentCore documentation.
-
Long-term and short-term memory – Deliver more personalized and relevant interactions by equipping agents with a memory system to recall context from current conversations and long-term knowledge. For more information, see Get started with AgentCore Memory in the AgentCore documentation.
-
Simplified tool development and integration – Enable your agents to discover and use tools through a single, secure endpoint. Quickly turn your existing enterprise resources into agent-ready tools with just a few lines of code, freeing developers to focus on building unique capabilities. For more information, see Get started with AgentCore Gateway in the AgentCore documentation.
-
Secure and scalable infrastructure – AgentCore provides a secure and scalable environment for deploying and operating agents. It includes features for identity and access management, data encryption, and network security. For more information, see Get started with AgentCore Identity in the AgentCore documentation.
-
Integration with a wide range of tools – Allows you to integrate your agents with a variety of tools, including a code interpreter and a browser tool that you can build by using the AgentCore built-in tools. For more information, see Get started with AgentCore Code Interpreter and Get started with AgentCore Browser in the AgentCore documentation.
-
Comprehensive observability and monitoring – Gain deep visibility into your agents with comprehensive tools to trace, debug, and monitor their performance in production. Visualize the agent's entire execution path to audit its reasoning and resolve failures. Use real-time dashboards and standardized telemetry data to track key operational metrics. For more information, see Add observability to your Amazon Bedrock AgentCore resources in the AgentCore documentation.
When to use AgentCore
AgentCore is particularly well-suited for autonomous agent scenarios including:
-
Organizations that want to accelerate development and lower operational overhead with a fully managed service that handles infrastructure, security, built-in tools, observability and scaling
-
Projects that need flexibility with modular services that work together or independently and are compatible with any framework, like CrewAI or LangGraph, and any foundation model from any source
-
Use cases that require stateful, conversational agents that need to maintain context and learn from past interactions to provide personalized and relevant responses
-
Agents enabled to perform complex tasks through simple integration with diverse applications, data sources, and APIs
Implementation approach for AgentCore
AgentCore is designed for organizations who want to move AI agents from proof of concept, built using open source or custom agent frameworks, to production. With AgentCore, organizations can do the following:
-
Deploy agents securely on serverless infrastructure, supporting any framework and model, with session isolation and built-in identity and access management for end-to-end security and compliance. Quickly create AgentCore Runtime agents for leading agent frameworks by using the starter toolkit.
-
Enhance agents by integrating persistent memory for context retention, simplifying tool development and integration through AgentCore Gateway. Leverage built-in browser tool and code interpreter for advanced workflows.
-
Trace, debug, and monitor AI agents in production using observability dashboards powered by Amazon CloudWatch Application Insights and OpenTelemetry, tracking key metrics of AgentCore resources (runtime, memory, gateway, and tools).
-
Accelerate deployment and innovation with fully managed, modular services, composable blocks together or independently, with any agent framework and model provider. This flexibility helps organizations move from prototype to production faster.
This managed approach enables organizations to quickly and securely build, deploy, and run enterprise-grade AI agents and multi-agent systems at any scale.
Real-world example of AgentCore
AWS has observed that one of Latin America's largest banks has used AI/ML for years to deliver a hyper-personalized and secure digital banking experience. The bank is expanding the agentic AI services by using AgentCore to provide customers with intuitive interactions, enhanced security, and greater automation. According to the CTO, AgentCore is expected to support their efforts to meet customer commitments at scale. AgentCore provides their developers the tools and flexibility to build and manage agents, while helping to ensure compliance with financial regulations.
