AWS Bedrock AgentCore
Amazon Bedrock AgentCore Samples is a repository of examples and tutorials for deploying and operating AI agents securely at scale using Amazon Bedrock AgentCore. It is framework-agnostic and model-agnostic, supporting frameworks like Strands Agents, CrewAI, LangGraph, and LlamaIndex, and any LLM. It is for developers who want to productionize agentic AI applications without building specialized agent infrastructure.
🎬 Building your first production-ready AI agent with Amazon Bedrock AgentCore | AWS Show & Tell · AWS Events
✨ Key features
- Framework-agnostic and model-agnostic agent deployment
- Secure, serverless runtime for agents and tools
- Gateway to convert APIs and Lambda into MCP tools
- Managed memory for personalized agent experiences
- Built-in tools: Code Interpreter, Browser, Web Search
- Observability with OpenTelemetry and evaluation support
🎯 Use cases
- Deploy a LangGraph agent to production on AWS
- Add managed memory to a CrewAI agent
- Expose existing APIs as MCP tools via Gateway
- Run LLM-as-a-Judge evaluations on agent responses
- Build a full-stack agentic application with identity and observability
📦 Installation
🧰 Requirements: Requires an AWS account with credentials configured, Node.js 20.x or later, uv for Python agents, model access to Anthropic Claude 4.0 in Amazon Bedrock, and AWS permissions BedrockAgentCoreFullAccess and AmazonBedrockFullAccess.
npm install -g @aws/agentcoreagentcore create
cd my-agent🚀 Usage
agentcore devagentcore deployagentcore invoke⚠️ Good to know
The Starter Toolkit CLI is no longer supported; use the AgentCore CLI instead. Some samples are legacy and will be updated.
❓ FAQ
What is the recommended way to create and deploy agents?
The AgentCore CLI (@aws/agentcore) is the recommended tool. Install it with npm install -g @aws/agentcore and use commands like agentcore create, agentcore dev, and agentcore deploy.
Which frameworks are supported?
AgentCore is framework-agnostic and supports Strands Agents, CrewAI, LangGraph, LlamaIndex, and others. The CLI supports Strands, LangGraph, LangChain, Google ADK, OpenAI Agents, and BYO.
What are the prerequisites for using these samples?
You need an AWS account with credentials, Node.js 20.x or later, uv for Python agents, model access to Anthropic Claude 4.0 in Amazon Bedrock, and the AWS managed policies BedrockAgentCoreFullAccess and AmazonBedrockFullAccess.
How can I add capabilities like memory or identity to my agent?
Use the AgentCore CLI commands: agentcore add memory, agentcore add identity, agentcore add evaluator, and agentcore add online-eval, then deploy with agentcore deploy.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.