Runtime, memory, identity, gateway, observability — agents as deployed services.

Module 11 of 16 · track Ship · Estimated time: 8–10 hours · Prerequisites: Modules 05–06. Module 08 for the LangGraph deployment.

Everything so far ran in a notebook. AgentCore is how an agent becomes a service with an identity, a memory store, a gateway, traces and a bill. This module deploys agents three ways — Strands, LangGraph, and no framework at all — so you can see the runtime is framework-agnostic.

Open the module on GitHub ↗: the slides, notebooks, exercises and solutions are there. This page is the reading order, with the tutorial and the workbench alongside.

The lesson that frames this module is P3 Run & Learn: How to Run an AI Agent in Production Safely (7 min): what running an agent as a service asks of the team. Then try the two-number report builder in the workbench: the same decision, with your numbers.

• Deploy an agent to AgentCore Runtime and invoke it as a service
• Wire AgentCore Memory and reason about retention
• Configure identity and tool access with least privilege
• Route traffic through the gateway and read the traces
• Estimate runtime cost and capacity before you commit

Portable GenAI concepts	AWS-specific surface
Agent as a service	Bedrock AgentCore Runtime
Session and identity	AgentCore Memory
Observability	AgentCore Identity
Harness engineering	AgentCore Gateway
	AgentCore Observability
	CloudWatch
	IAM
	CDK

Portable concepts transfer to any stack. The AWS column is where this module touches the cloud — see docs/concepts/portability-matrix.md.

#		Step	What it is
1	📖	slides/AgentCore_Part1_Foundations.pptx	AgentCore foundations
2	💻	notebooks/00_setup_and_sanity.ipynb	Setup and sanity check
3	📖	walkthroughs/01_agentcore_foundations.md	Foundations walkthrough
4	💻	notebooks/01_build_an_agent_three_ways.ipynb	One agent, three ways
5	💻	notebooks/02_agentcore_runtime.ipynb	Runtime
6	💻	walkthroughs/AgentCore_01_Strands_Minimum_Deploy.ipynb	Minimum deploy — Strands
7	💻	walkthroughs/AgentCore_02_NoFramework_Minimum_Deploy.ipynb	Minimum deploy — no framework
8	💻	walkthroughs/AgentCore_03_LangGraph_Minimum_Deploy.ipynb	Minimum deploy — LangGraph
9	💻	notebooks/03_memory.ipynb	Memory
10	💻	notebooks/04_tools_and_identity.ipynb	Tools and identity
11	✏️	exercises/exercise_A_parcelpilot.md	Exercise A — ParcelPilot
12	📖	slides/AgentCore_Part2_Advanced.pptx	AgentCore advanced
13	💻	notebooks/05_multi_agent_orchestration.ipynb	Multi-agent orchestration
14	📖	walkthroughs/03_harness_engineering.md	Harness engineering
15	✏️	exercises/exercise_B_voltdesk.md	Exercise B — VoltDesk
16	💻	notebooks/07_production_patterns.ipynb	Production patterns
17	📊	activities/AgentCore_Cost_and_Capacity_Workbench.xlsx	Cost and capacity workbench
18	✏️	exercises/exercise_END_renewq.md	Final exercise — RenewQ

📖 read   💻 run   ✏️ practise   📊 workbook   🔖 reference

These exercises are graded against a rubric rather than an answer key. Each one is a fresh domain (not TravelMind) with layered Base / Stretch / Advanced targets, and each ends with:

• a self-check rubric scored out of 100, so you can grade your own work honestly
• a pass/fail LLM-integrated task that is required regardless of how far you got
• facilitator and TA notes in an appendix, if you are running this for a group

README_exercises.md explains how the three map onto the notebooks.

src/MyFirstRuntimeAgent is a complete CDK-managed AgentCore project — app code, infrastructure, and configuration.

• Deploying without a memory retention decision. Storage grows, and so does the bill.
• Over-broad IAM on the runtime role. Scope tools to what the agent actually needs.
• Skipping observability until something breaks in production.

Every error anyone has hit in this course is in the Error Index, by the exact string. Model access and region problems are in Model and Region Notes. If it is a question rather than an error, check the answered ones first, then ask in Q&A.

Frameworks

• Blast Radius Grid — Scope identity per tool, not per agent
• Cost Cliff Map — What bills while it exists, and what bills per use
• Reversibility Test — Four things must roll back together

Quick reference

• AgentCore — The five primitives
• IAM for agents — Least privilege that actually holds
• Observability — The three vital signs

Recipes and procedures

• Deploy any agent to AgentCore — Framework-agnostic deploy path

---

← Module 10 · RAG, OpenSearch and LiteLLM · All modules · Module 12 · A2A and A2UI: Agent Interoperability →

_{✏️ This page is generated from modules/11-bedrock-agentcore/README.md. To change it, edit the module README: an edit made here is replaced at the next sync.}