Amazon Bedrock AgentCore Immersion Days
Hands-on labs that go with the Amazon Bedrock AgentCore series. Each immersion day is a step-by-step walkthrough; clone its repository to run the labs.
| # | Immersion day | Covers | Repository |
|---|---|---|---|
| 02 | Harness | The AgentCore harness is a managed agent loop that you set up with configuration instead of code. | github.com/chiwaichan/agentcore-immersion-day-02-harness |
| 03 | Runtime | AgentCore Runtime hosts your agent code and gives every session its own microVM. | github.com/chiwaichan/agentcore-immersion-day-03-runtime |
| 04 | Memory | AgentCore Memory gives an agent two kinds of memory: the raw turns of a conversation, and the lasting knowledge extracted from them. | github.com/chiwaichan/agentcore-immersion-day-04-memory |
| 05 | Gateway | AgentCore Gateway is one managed MCP endpoint that sits in front of many back-end tools. | github.com/chiwaichan/agentcore-immersion-day-05-gateway |
| 06 | Identity | AgentCore Identity gives an agent an identity of its own and the credentials to act for a user, without the agent code ever holding a client secret. | github.com/chiwaichan/agentcore-immersion-day-06-identity |
| 07 | Payments | AgentCore Payments lets an agent pay for what it uses, within a budget it cannot raise itself. | github.com/chiwaichan/agentcore-immersion-day-07-payments |
| 08 | Code Interpreter | AgentCore Code Interpreter gives an agent an isolated sandbox in which to run code it wrote a moment ago. | github.com/chiwaichan/agentcore-immersion-day-08-code-interpreter |
| 09 | Browser | AgentCore Browser gives an agent a managed web browser that runs in the cloud, one isolated microVM per session. | github.com/chiwaichan/agentcore-immersion-day-09-browser |
| 10 | Observability | AgentCore Observability puts an agent's traces, logs and metrics into Amazon CloudWatch, wherever the agent runs. | github.com/chiwaichan/agentcore-immersion-day-10-observability |
| 11 | Policy | AgentCore Policy puts deterministic rules in front of an agent's tools. | github.com/chiwaichan/agentcore-immersion-day-11-policy |
| 12 | Evaluations | AgentCore Evaluations scores what an agent actually did, from the traces it writes to Amazon CloudWatch. | github.com/chiwaichan/agentcore-immersion-day-12-evaluations |
| 13 | Optimization | AgentCore Optimization improves an agent from its own traces: it recommends a better system prompt or better tool descriptions, keeps configuration in versioned bundles the agent reads at run time, and A/B tests a new configuration on live traffic before you ship it. | github.com/chiwaichan/agentcore-immersion-day-13-optimization |
| 15 | Security | This is the threat-driven day. You deploy a small procurement agent to AgentCore Runtime the "first draft" way, run a posture check that lists what is wrong with it, and then close each gap against a named threat, ending with a posture report that has no FAILs. | github.com/chiwaichan/agentcore-immersion-day-15-security |