Use your IAM credentials to access the AgentCore console. From the AWS Management Console, navigate to Bedrock AgentCore to begin your setup.

AgentCore is modular. You can start with Runtime, Memory, Gateway, Tools, Identity, or Observability — choose based on your use case. There's no required order or dependency.

Fill in the required settings for your selected service: Upload a container image for Runtime to deploy your agent, define tool targets in Gateway to connect your services, create a store in Memory to maintain context, or set up authentication rules in Identity to manage access.

Use the console to test your runtime or tools, invoke agents, simulate calls, and monitor logs and traces. Observability can be enabled to inspect workflows and performance metrics.

Choose between three programmatic access methods: AWS SDK (software development kits in Python, Java, etc. for application integration), AWS CLI (command-line interface for quick terminal commands), or REST APIs (direct HTTP requests). These options let you interact with AgentCore services—such as Runtime, Memory, Gateway, Identity, Tools, and Observability—through the workflows that best match your development environment.

Configure authentication using AWS credentials: IAM roles, access keys, or OAuth tokens. Each AgentCore service may have specific requirements, especially when integrating with third-party tools or defining identity scopes. Refer to the documentation to ensure secure setup across services.

Use your preferred method to provision AgentCore services. For example, create a runtime agent using the SDK with agentcore_runtime.create_agent(), register a tool with agentcore_gateway.register_tool(), or create a memory store via CLI using aws agentcore create-memory-store --name MyMemoryStore --type VECTOR. Each service supports focused configuration to fit your use case.

Trigger agent execution using InvokeAgent or related SDK methods. Monitor performance with AgentCore Observability and OpenTelemetry pipelines. These tools give you visibility into both individual service performance and complete agent workflows.

Select either an Agentic IDE like Kiro, or an AI coding assistant such as Claude Code, GitHub Copilot, or Q Developer CLI. Both types of tools work seamlessly with the AgentCore MCP Server to provide natural language interfaces for AgentCore development.

Install the AgentCore MCP Server with a single command to create a bridge between your chosen development environment and AgentCore services, enabling natural language workflows.

Your Agentic IDE or AI coding assistant helps you connect AWS credentials and configure project settings. The MCP Server handles AgentCore service configurations and dependencies in the background.

Use natural language commands within your development environment to transform code, deploy agents, manage tools, and troubleshoot. The MCP Server translates your intentions into proper AgentCore operations.