Building Multi-Agentic Workflows on AWS Bedrock + AgentCore

Name: Building Multi-Agentic Workflows on AWS Bedrock + AgentCore
Rating: 4.4 (258 reviews)

Create Multi-Agent Workflows, Deploy on AgentCore + Build Memory-Enabled Strands Agents + DynamoDB and Web Search Tools

Created byPatrik Szepesi

Last updated 8/2026

English

What you'll learn

Understand and Implement Multi-Agent Workflows
Deploy Multi-Agent Workflows with AWS- using Bedrock, Lambdas, API Gateway, S3 and many more
Add long-term memory with semantic and preference strategies to your Agents
Leverage AWS Bedrock for LLMs
Deploy agents to production on AgentCore Runtime
Create agents that remember users across sessions
Implement Multi-Agent Collaboration
Deploy Production AI Systems – Set up a scalable AI architecture using AWS Lambda and API Gateway.
Create Action Groups in AWS Lambda – Build and manage action groups for AI decision-making in serverless environments.
Build AI-Powered Travel Agents – Design an intelligent travel assistant that can provide accommodation and restaurant recommendations.
Understand short-term vs long-term agent memory
Understand the Pricing for Bedrock AgentCore Runtime
Implement API Gateway for External Access – Expose your AI travel agent to the web using AWS API Gateway.
Optimize AI Requests with API Rate Limits – Learn how to manage API request limits and prevent excessive usage costs.
Implement Logging and Monitoring – Track AI model performance and monitor API usage with AWS CloudWatch.
Understanding the Pricing for Bedrock Agentcore Long and Short Term Memory
Understand the Role of Supervisor Agents – Learn how supervisor agents manage and coordinate tasks efficiently.
Deploy an End-to-End AI System – Take your travel agent from concept to production in a real-world AWS environment.
Fine-Tune AWS Bedrock LLM Responses – Adjust system parameters to improve the accuracy and relevance of travel recommendations.
Design Scalable Serverless Applications – Learn best practices for scaling AI-driven serverless applications in AWS.
Build web search agents with Strands and DuckDuckGo
Implement short-term memory for conversation tracking
Use lifecycle hooks to load and save agent memory
Build a personal assistant with Claude Haiku and live web search
AWS Strands Framework

Course content

14 sections • 87 lectures • 11h 49m total length

What the Course Covers6:15

Signing In to AWS4:42
Set up your AWS account, sign in to the console using the root account, enable multi-factor authentication, and create IAM user roles to safely manage costs and security.
Creating and IAM User in AWS5:45
Create an IAM user in AWS with management console and CLI access, an auto-generated password that must be changed on first sign-in, and follow least-privilege guidance.
Accessing Our AWS Environment with our IAM User3:23
What to Do If your AWS Account Gets Hacked1:32
Learn how to respond to a hacked AWS IAM user by signing in with the root account, deleting the compromised user, and restoring secure access for your AWS Bedrock workflows.

What We Are Building Part 15:37
Learn to build production‑ready multi‑agent workflows on AWS bedrock, deploying via API gateway, with a travel supervisor coordinating accommodation and restaurant agents using S3 data.
What We Are Building Part 25:00
Explore how a supervisor agent coordinates restaurant lookups with a restaurant agent in a serverless AWS Bedrock workflow, using pandas and Lambda for precise queries.

Introduction to Bedrock Foundation Models13:30
Deep Dive Into Inference Configurations26:37
Inference Profiles, Model Catalog, Provisioned Throughput and More Theory17:18
Prompt Management, Optimization, BDA, AgentCore Intro and more18:50
Exploring the Playground in AWS Bedrock16:19
Exploring Modal Providers, Modalities, API Invocation and Pricing15:31
Quotas, Model Comparison, Guardrails, and More27:33
Image Generation in the Playground with Diffusion Models2:56
Prompt Management with AWS Bedrock15:43
Intelligent Prompt Routing with AWS Bedrock10:52
Using LLMs in Batch Inference Mode in AWS Bedrock13:01

Introduction to AWS Bedrock Agents1:04
Explore using AWS Bedrock builder tools to create supervisor, accommodation, and restaurant agents, and prepare to set up these agents from scratch in Bedrock.
Creating the Restaurant Agent18:25
Configure the restaurant agent in AWS Bedrock, enable multi-agent collaboration for the supervisor agent, and set up a List restaurants lambda with city and fine dine parameters.
Creating our AWS S3 Bucket To Store Our Data4:10
Create an s3 bucket to store hotels, airbnbs, and restaurants csvs for your bedrock workflow, then upload the restaurants csv with name, city, and fine dining status.
Uploading Restaurant Data to AWS S30:47
Upload the restaurant CSV to an AWS S3 bucket, confirming the file is in the bucket and ready for the restaurant agent's Lambda function in the next video.
Creating an Action Group For Our Restaurant Agent13:51
Create a lambda function for the bedrock restaurant agent, configuring S3 bucket and restaurant.csv, extract parameters into a dictionary, and enable a pandas layer to read the csv.
Finishing Our Lambda Function for our Restaurant Agent14:18
Finish implementing the restaurant agent's lambda function by loading a CSV into a dataframe, normalizing city and fine dining fields, applying filters, and returning JSON with error handling.
Testing Our Restaurant Agent15:36
Test the restaurant agent in AWS Bedrock by interacting with the lambda-backed service, reviewing configurations, tracing steps, fixing S3 access, and debugging common errors, including timeout settings.

Setting Up The Accommodation Agent9:00
Design and deploy an AWS Bedrock accommodation agent that selects hotels or Airbnbs using location, pets, pool, and sauna filters, with lambda functions and S3 data.
Uploading Our Hotel and Airbnb Data to AWS S31:14
Upload hotel and Airbnb csv files to the same s3 bucket as the restaurants csv to prepare data for refining the accommodation agent Lambda function.
Creating The Lambda Function Action Group For The Accommodation Agent19:02
Build a lambda function action group for the accommodation agent that reads hotel and Airbnb csvs from s3 and applies location, pets, pool, and sauna filters.
Finishing The Accommodation Agent10:55
Finish the accommodation agent by refining a pandas-based lambda to normalize data, apply dynamic filters, and output JSON for the Bedrock model, with S3 access and deployment steps.
Testing The Accommodation Agent9:00
Test and debug the accommodation agent in bedrock, validating hotel and Airbnb searches, pet-friendly options, pools and saunas, and the supervisor agent orchestration.

Creating and Testing The Supervisor agent9:38
Create and test a supervisor agent that orchestrates multi-agent collaboration with restaurant and accommodation collaborators on AWS Bedrock, enabling history sharing, precise routing, and final responses.
Explaining Agent Collaborators4:52
Clarify how collaborator instructions guide the supervisor agent to route to accommodation or restaurant agent. Avoid treating collaborators as separate agents; provide detailed instructions to improve accuracy and reduce hallucinations.
Multi Agent UI Enhancement, Timing Agents3:01
Observe how the enhanced UI in Bedrock supervisor demonstrates multi-agent collaboration with a trace timeline, revealing when the supervisor, restaurant, and accommodation agents run and for how long.

Requirements

Basic Python Programming

Description

Want to build AI applications where multiple agents collaborate, remember users, and run in production? This course takes you from multi-agent fundamentals to deploying intelligent, memory-enabled agents on AWS Bedrock and AgentCore.

You'll build a fully operational travel planner where Supervisor Agents coordinate tasks while Collaborator and Helper Agents handle database lookups, API calls, and travel preferences on your behalf. You'll also build a personal assistant agent with live web search powered by DuckDuckGo — capable of fetching real-time information and responding with up-to-date answers.

What You'll Learn:

Multi-Agent Design — When to break tasks into specialized agents, how to handle inter-agent communication, and how to ensure seamless collaboration
AWS Bedrock LLMs — Customize prompt templates, override parameters, and optimize AI output using foundation models
Serverless Deployment — Store data in S3, build with Lambda Action Groups, and deploy via API Gateway for live, scalable requests
AgentCore Runtime — What Amazon Bedrock AgentCore is and how to deploy and run agents at scale on purpose-built infrastructure
Web Search Agents — Build agents using the Strands framework with Claude Haiku that search the web in real time via DuckDuckGo
Short-Term Memory — Track conversation context within a session using AgentCore's get_last_k_turns
Long-Term Memory — Configure extraction strategies that automatically capture Semantic facts, User Preferences, and Session Summaries — so your agents remember users across sessions
Understanding the AWS agent harness and runtime architecture
Swarm-based agent collaboration
Graph-based workflows
Agents as Tools
Agent-to-Agent communication

By the End of This Course, You Will Be Able To:

Orchestrate Supervisor, Collaborator, and Helper Agents for real-world scenarios
Deploy agents on AgentCore Runtime with production-grade infrastructure
Build agents that search the web and respond with live information
Give agents short-term and long-term memory that persists across sessions
Deliver dynamic, personalized recommendations powered by multi-agent AI
Choosing the right multi-agent pattern for your use case

Whether you're an aspiring AI developer or a seasoned engineer — this course gives you the hands-on skills to build agents that don't just respond, but remember, personalize, and improve over time. Join us and start building the next generation of AI with AWS Bedrock and AgentCore.

Who this course is for:

AI and Machine Learning Engineers who want to leverage Large Language Models (LLMs) and advanced agent orchestration techniques.
Cloud Architects and Engineers looking to design robust, serverless AI architectures on AWS with production-grade scalability.
Data Scientists interested in expanding their skill set to include multi-agent systems and real-time deployment in the cloud.
DevOps and MLOps Professionals eager to master end-to-end automation for AI applications, from data storage to API endpoints.
Software Developers keen on integrating conversational AI into their applications for dynamic, user-friendly experiences.
Tech Enthusiasts and Entrepreneurs who see the potential of AI-driven services and want a hands-on approach to rapid prototyping and scaling.

Building Multi-Agentic Workflows on AWS Bedrock + AgentCore

What you'll learn

Explore related topics

Course content

Course Overview1 lecture • 6min

Setting Up Our AWS Account4 lectures • 15min

Multi Agentic Capstone Project2 lectures • 11min

(Optional) Introduction to AWS Bedrock11 lectures • 2hr 58min

Course Resources1 lecture • 1min

Architecture Design1 lecture • 7min

Multi Agentic Setup Past 2026 July 31. AWS Strands Update1 lecture • 1min

Creating the Restaurant Agent in AWS Bedrock7 lectures • 1hr 8min

Creating the Accommodation Agent in AWS Bedrock5 lectures • 49min

Multi Agentic Collaboration Using the Supervisor Agent in AWS Bedrock3 lectures • 18min

Requirements

Description

Who this course is for: