agentcore-samples

Why we included this project

The agent systems people actually run in production have more moving parts than a single chat loop, and this repo is a working catalog of those parts. It is a set of runnable samples for AgentCore, Amazon Bedrock's managed agent runtime, aimed at teams building with LangGraph, CrewAI, LlamaIndex, Strands, or their own stack. The examples cover the pieces that trip people up once a demo turns into a deployment: identity and access control, managed memory, turning APIs into MCP-compatible tools, and OpenTelemetry-based observability. The getting-started exercises run through the AgentCore CLI, while feature folders, end-to-end applications, and infrastructure-as-code templates show how an agent slots into a real AWS stack. If you are moving an agent out of a notebook and onto AWS infrastructure, this is a practical map of what the managed runtime handles for you and which parts you still wire up yourself.