Build, deploy and operate agents with AgentCore
Three scenarios, one morning. Each one adds a different part of Amazon Bedrock AgentCore.
This is your guide for the morning. Keep it open next to the AWS console and your code editor. Every command you need is on these pages with a Copy button, and every step has a Done tick so you can find your place again if you step out. Your ticks are saved in this browser only.
What you'll do
You'll work through three short business scenarios. Each one has a problem an agent can help with, and each one adds a different part of Amazon Bedrock AgentCore. By the end you'll have used the main pieces and you'll know which ones you would reach for in your own stack.
| Scenario | What you add | Where the agent runs | |
|---|---|---|---|
| 1 | Kopi Run. Office kopi orders in plain English | A harness agent, Gateway, Memory | On AgentCore |
| 2 | Tally. A billing support agent that gives away credits | Gateway, Policy, Observability | In your terminal, standing in for ECS, a VM or another cloud |
| 3 | Intake. A finance agent that repeats a bad number | Evaluations, Code Interpreter | On AgentCore |
Scenario 2 is the one to pay attention to if you already have an agent running somewhere. AgentCore doesn't need to host your agent for you to use the rest of it.
You won't be writing an agent framework from scratch. We provide the code and config, and your job is to read them, run them and understand what each piece does. An AI coding assistant is optional. Where one helps, there's a Kiro corner box with a prompt to try.
Agenda
| Time | Block |
|---|---|
| 08:30 | Doors open. Coffee, plus a setup desk by the entrance if you want a head start |
| 09:00 | Welcome, and AgentCore in fifteen minutes |
| 09:15 | 0. Setup and access |
| 09:45 | 1. Kopi Run |
| 10:20 | Break |
| 10:35 | 2. Tally |
| 11:35 | 3. Intake |
| 12:25 | Wrap-up and clean-up |
| 12:40 | Questions, then one-to-one consults with AWS and JAM |
How to use this guide
- Work in pairs if you like. One person drives and the other reads ahead.
- Each scenario opens with a short story. Read it, because it tells you what the agent is for.
- You can build in the AWS console or in a terminal. Pick a path during setup, and switch any time from the sidebar. Steps that differ have Console and Terminal tabs.
- Use the Copy button instead of retyping. Anything you need to change yourself looks like
<this>. - Green Done when boxes tell you what working looks like. If yours doesn't match, check the amber Stuck? box underneath before you put your hand up.
- Every scenario has a catch-up command that jumps you to the finished state. Feel free to use it whenever you want.
- Deploys take three to five minutes, so read the Under the hood boxes while you wait.
What you need
- A laptop with a current browser. Everything else, including the code editor, runs in a workshop AWS account we give you.
- The event access code, which you'll find on the screen at the front of the room.
After today
Your workshop account stays open for a while after the session (we'll tell you when it closes), so you can finish anything you didn't get to. The code is at https://github.com/jamadmin/agentcore-workshop.git and every command works the same way in your own AWS account in ap-southeast-1.
It's worth bookmarking this guide: https://agentcore-workshop.jam247.com/