AgentCore Managed Harness: Getting Started

Stop wiring up containers, storage, and orchestration code just to get an agent running. The managed harness in Amazon Bedrock AgentCore replaces all of that with a config file and a system prompt. In this video we'll show you how to build a real AI research agent that opens a managed browser, scrapes the top stories from Hacker News, pulls Reddit via RSS, saves snapshots to persistent storage, and generates a trend digest every 6 hours. The whole thing is defined in a JSON config, deployed from the CLI, and automated with EventBridge + Lambda. Then we'll wire it up to Telegram so the digest shows up on your phone. What you get out of the box with the managed harness: ✅ Isolated micro VM per session with its own shell and file system ✅ Persistent storage at /mount/data that survives restarts ✅ Cross-session memory scoped by actor ID ✅ Hot-swap models, tools, and prompts per API call — no redeploy ✅ Shell access via --exec (no token cost) ✅ Connects to browser, code interpreter, identity, observability Amazon Bedrock AgentCore Managed Harness News: https://go.aws/4cGjW6F Amazon Bedrock AgentCore Managed Harness Doc: https://go.aws/3QHzjDg Follow AWS Developers! 🆇 X: https://go.aws/4cL1ecI 💼 LinkedIn: https://go.aws/3PbmBMH 0:00 Introduction - Why Managed Harness? 0:25 Understanding Agent Architecture 2:45 Installing Agent Core CLI 4:21 Adding Resources 5:25 Adding the System Prompt 7:09 Running AgentCore Invoke 9:02 Lambda Function Setup 10:05 Telegram Integration Demo 11:10 Final Results & Wrap-up #AWS #AmazonBedrockAgentCore #ManagedHarness