What is Learn AgentCore?
Learn AgentCore is an interactive course for Amazon Bedrock AgentCore. Not a model. Not LangGraph. A platform for running agents in production: isolation, identity, tools, policy, memory that outlives a process.
You step a live session in the browser and watch the jobs after the loop works: Tuesday traffic, two customers on one host, Slack as Maya, a $500 rule. The goal is to get someone from "I have a graph on localhost" to "I can see what has to sit around that graph."
Why We Built It
AgentCore is easy to name and hard to place. The docs will give you Runtime, Gateway, Identity, Policy, Memory, Observability. The thing that actually matters is the job after the loop works.
LangGraph is the loop. AgentCore is how that loop runs when two customers share a host, when an IdP has to say who may call, when the company API should be wrapped once, when $500 is a rule, when terse replies must still be true next month, or when you need to find a denied refund in a trace.
We built Learn AgentCore so people can step that factory floor instead of only reading the product names.
A Hands-On Way to See the Session
This is not a slide deck about AWS.
You pick a module, press Step, and watch a session live: a microVM, inbound identity, outbound OAuth, Gateway MCP, Cedar policy, short-term and long-term memory, traces. No AWS key. Quizzes stay in this browser. Sign in with your En Dash account if you want progress and places on another machine.
Built Around Jobs, Then the Platform
Two doors.
Use cases are the jobs: Tuesday traffic, two customers, Slack as Maya, a $500 spend cap, memory that outlives the process, a denied refund you can find. Each one is "what localhost does" versus "what the platform does," then a link into the matching module.
Learn is twelve modules on the AgentCore platform itself: Runtime sessions, Harness, Gateway, Identity, Policy, Memory, Observability. The capstone puts them in one session.
Useful for Engineers Who Already Have a Loop
This is for people whose graph works on one laptop with a dotenv, and who now have a second customer.
Production isolation. An IdP. MCP wrap. Deterministic policy. Durable memory. Traces. If the agent still lives on localhost, skip it. Learn AgentCore is for the other case.
Built to Make the Idea Click
Some product ideas fit in one sentence. AgentCore needs a session you can watch.
Learn AgentCore exists so more people can see what the platform is for, when a laptop is cheaper, and what it looks like when an agent actually leaves.
Not official AWS. A community resource from En Dash, same shop as Learn LangGraph, html-in-canvas.dev, and learn-pretext.com.