Get started with AgentCore Runtime using the AgentCore CLI in TypeScript

This page provides a concise workflow for creating and deploying a code-based TypeScript Strands agent to AgentCore Runtime. For CLI installation, complete prerequisites, project capabilities, observability, and cleanup, see Get started with Amazon Bedrock AgentCore.

Prerequisites

Create your agent project

Create a TypeScript project that uses the Strands Agents SDK and Amazon Bedrock:

agentcore create \
  --name MyTsProject \
  --template agent-typescript-strands

The template creates a Node.js 22 CodeZip Runtime, adds an AgentCore Memory resource, and installs the project dependencies.

For other model providers and the CLI’s China Region restrictions, see Choose a Runtime model provider and CLI project scope in China Regions. TypeScript Strands supports the openai_compatible provider but not lite_llm.

You can also run agentcore with no arguments to open the interactive menu. Outside a project, the menu offers create. Inside a project, it offers deploy, invoke, and status. Choosing dev from the menu starts the local server in headless mode. To open the Agent Inspector, run agentcore dev --mode browser.

Test your agent locally

Start the Runtime locally and open the Agent Inspector:

cd MyTsProject
agentcore dev --mode browser

Without --mode browser, agentcore dev defaults to headless mode and streams the Runtime output to your terminal.

Build and deploy your agent

agentcore build
agentcore deploy

Check the deployed resources:

agentcore status

Invoke your deployed agent

agentcore invoke --runtime agent \
  --payload '{"prompt":"Hello, what can you do?"}'

Iterate on app/agent/main.ts, test locally, and deploy again when you are ready.

Troubleshoot

For common CLI installation, deployment, configuration, and observability issues, see Troubleshoot.

Next steps