This folder contains Hands-on tutorials for building, deploying, and managing AI agents with Amazon Bedrock AgentCore.
AgentCore services work independently or together, with any agentic framework (Strands Agents, LangChain, LangGraph, CrewAI, etc.) and any model.
- An AWS account with Amazon Bedrock access
- Python 3.10+ and Jupyter Notebook (or JupyterLab)
- AWS CLI configured with appropriate credentials
- Basic familiarity with AI agents and AWS services
01 - Runtime
Deploy and scale AI agents on a secure, serverless runtime -- regardless of framework, protocol, or model. Covers hosting agents, MCP servers, A2A, and bi-directional streaming. (Docs · Deep Dive Video)
02 - Gateway
Turn APIs, Lambda functions, and existing services into MCP-compatible tools without managing integrations. Includes examples for auth, access control, sensitive data masking, and more. (Docs · Deep Dive Video)
03 - Identity
Manage agent identity and access across AWS services and third-party apps (Slack, Zoom) using standard identity providers (Okta, Entra, Cognito). Covers inbound auth, outbound auth, and 3LO flows. (Docs · Deep Dive Video)
04 - Memory
Add fully managed memory to your agents for personalized experiences. Explore short-term memory, long-term memory, branching, and security patterns. (Docs · Deep Dive Video)
05 - Tools
Use AgentCore's built-in tools: Code Interpreter for secure code execution, and Browser Tool for web navigation and form completion. (Code Interpreter Docs · Browser Tool Docs · Deep Dive Video)
06 - Observability
Trace, debug, and monitor agent performance with OpenTelemetry-compatible telemetry. Works for agents hosted on Runtime, self-hosted agents, Lambda-based agents, and EKS-hosted agents. (Docs · Deep Dive Video)
07 - Evaluations
Assess agent quality with built-in and custom evaluators across dimensions like correctness, helpfulness, and safety. Includes creating evaluators, running evaluations, and using results. (Docs · Deep Dive Video)
08 - Policy
Define and enforce security controls using Cedar language policies to prevent data leakage and authority overreach. Covers natural language policy authoring and fine-grained access control. (Docs · Deep Dive Video)
09 - End-to-End Workshop
Build a complete agent step by step, combining Runtime, Gateway, Identity, Memory, and more into a production-ready solution. (Deep Dive Video)
- New to AgentCore? Start with 01 - Runtime and work through the tutorials in order.
- Looking for a specific capability? Jump directly to any tutorial -- each one is self-contained.
- Want the full picture? The End-to-End Workshop ties all the components together.
- Amazon Bedrock AgentCore Documentation -- Official developer guide and API reference
- AgentCore Deep Dives Playlist -- Video playlist covering each component in detail