Deploying Strands Agents to Amazon Bedrock AgentCore Runtime
Amazon Bedrock AgentCore Runtime is a serverless runtime for deploying and scaling agents built on any framework, including Strands Agents, LangChain, LangGraph, and CrewAI. It carries any protocol (MCP, A2A) and any model from any provider (Amazon Bedrock, OpenAI, Gemini), and runs multi-modal, real-time, and long-running agents alike.
Each user session gets its own dedicated microVM, so state and privileged operations stay isolated between users. Sessions persist, the runtime scales to thousands of concurrent sessions in seconds, and you pay only for actual usage rather than provisioned infrastructure. Through AgentCore Identity, it integrates with identity providers such as Amazon Cognito, Microsoft Entra ID, and Okta, and OAuth providers such as Google and GitHub, so you authenticate with OAuth tokens, API keys, or IAM roles without building custom security infrastructure.
Prerequisites
Section titled “Prerequisites”Before you start, you need:
- An AWS account with appropriate permissions
- Python 3.10+ or Node.js 22+
- Optional: A container engine (Docker, Finch, or Podman) - only required for local testing and advanced deployment scenarios
Choose Your Language
Section titled “Choose Your Language”Pick the language your agent is written in to continue: