Skip to content

Getting Started with the AI Foundation Platform

Welcome! 👋

This guide is for complete beginners.

  • Never touched an AI agent before? Good.
  • Never heard of "MCP"? Also good.
  • New to backend development in general? You're in the right place.

We'll explain every term the first time we use it. No prior knowledge needed.

By the end of this guide, you will have built and launched your very own AI agent — using only clicks in a web app. No coding required.

What you'll need before starting

Before you begin, make sure you have:

  • [ ] Access to the AI Foundation Management Platform. This is the web app you'll click through for this whole guide.
  • [ ] A login for that platform, with permission to create and deploy agents. (If you don't see a "+ Create Agent" button once you log in, ask a platform admin to grant you access.)
  • [ ] About 20–30 minutes of uninterrupted time.

That's it! You do not need:

  • Any coding experience.
  • Any command-line tools.
  • Any API keys or tokens.

Everything happens by clicking around a web page, just like using any other website.

What you'll accomplish

By the end of this guide, you will have:

  1. Created an agent — an AI assistant you configure yourself.
  2. Registered a tool — given your agent a new skill, like looking something up.
  3. Chatted with your agent — to see it actually respond to you.
  4. Deployed your agent — put it "live," so it can be used outside your own screen.

Don't worry if any of that sounds unfamiliar right now. We'll explain each idea, one small step at a time, as you go. You've got this!

Key Concepts

Here are a few words you'll see over and over in this guide. We explain each one in plain English, with no jargon.

  • Agent — think of this as a little AI assistant you build yourself. You give it a name, tell it how to behave, and (optionally) give it some tools. There are two kinds:

    • A Simple agent — works entirely on its own. This is what you'll build in this guide.
    • An Orchestrator — a "manager" agent that hands off tasks to other agents (its "team") and combines their answers. You can explore this later.

  • MCP (Model Context Protocol) — a fancy name for a simple idea: a standard way for an agent to connect to outside tools and information. You don't need to understand how it works internally — just know that it's the plumbing that lets agents "reach out" and do things.

  • MCP Server — a service that offers a set of tools an agent can use. For example, one MCP Server might offer a "search our documents" tool. You register an MCP Server once, and then any agent on the platform can be given access to its tools.

  • Tool — one specific thing an agent can do, like "search the web" or "look up a customer order." Tools usually come from an MCP Server you've registered.

  • Deployment — this means making your agent "go live" — reachable in a real environment, instead of only testable on your own screen. Think of it like the difference between rehearsing a play in your living room versus performing it on a real stage where an audience can watch.

What you'll build

A small "hello world" agent — simple on purpose. It will have:

  • One tool it can use.
  • A chat conversation you test yourself.
  • A real deployment to AWS (Amazon's cloud), so it's genuinely live.

It's a tiny example. But the steps you follow here are the exact same steps you'll use later to build bigger, real production agents. Learn it once, use it forever.

How this guide is organized

Follow these five pages in order. Each page builds on the one before it.

Page What you'll do
1. Create an Agent Fill in the "+ Create Agent" form
2. Register a Tool (MCP Server) Add an MCP Server and pick which tools your agent can use
3. Test the Agent Chat with your agent from the home page or the Playground
4. Deploy the Agent Deploy your agent to AWS from its card
5. MCP SDK Reference Optional — for later, if you want to skip the UI and talk to the platform directly

Ready? Start with 1. Create an Agent.