AI Agentic Design Patterns with AutoGen
by DeepLearning.AI × Microsoft Research
Learn the four agentic design patterns from AutoGen's own creators, one working multi-agent project at a time.
Overview
Chi Wang, then Principal Researcher at Microsoft Research, and Qingyun Wu, Assistant Professor at Penn State — the two researchers behind AutoGen — teach this course, and it is structured one pattern per project rather than one API per lesson. Lesson 2 introduces two-agent conversation through a stand-up comedy routine; lesson 3 covers sequential chats and context carry-over via a customer-onboarding flow; lesson 4 demonstrates reflection, where a writer agent's draft is critiqued and revised by a reviewer agent; lesson 5 covers tool use through conversational chess, where a legal-move tool constrains what the agents can do; lesson 6 gives agents a code executor to perform real financial analysis; and lesson 7 combines planning with a group of agents producing a stock report. Each of those six ships a runnable notebook, and a graded quiz closes the course. One caveat the course itself will not tell you: AutoGen entered maintenance mode in 2026, receiving only bug and security fixes, and Microsoft now directs new projects to the Microsoft Agent Framework, which merges AutoGen with Semantic Kernel and reached 1.0 general availability in April 2026. The design patterns taught here are framework-independent and remain the reason to spend 95 minutes on it; the specific AutoGen-era API calls in the notebooks are not where you should start a new build.
At a Glance
- Topic
- Agentic
- Level
- Intermediate
- Format
- Course
- Cost
- Free
- Duration
- 1 hr 35 min, self-paced — 8 video lessons, 6 code notebooks, 1 graded quiz
- Provider
- DeepLearning.AI × Microsoft Research
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Implement reflection, where a reviewer agent critiques and improves another agent's draft
- ✓Build two-agent and sequential-chat conversations that carry context forward
- ✓Give agents tools and constrain their actions to legal moves
- ✓Attach a code executor so agents can run real analysis themselves
- ✓Coordinate a planner agent with a group of specialist agents
- ✓Use nested chats and summarisation between stages of a conversation
- ✓Judge which agentic pattern actually fits a given task
Highlights
- •Taught by AutoGen's two creators, Chi Wang and Qingyun Wu
- •Organised by design pattern, so the lessons outlive the framework
- •Six runnable notebooks, each a distinct working multi-agent system
- •Free and 95 minutes — a cheap way to evaluate the patterns before committing
- •Honest caveat: AutoGen is in maintenance mode, superseded by Microsoft Agent Framework
Who It’s For
Best For
- ✓Developers deciding which multi-agent pattern their problem actually needs
- ✓Engineers who learn architecture faster from working code than from diagrams
- ✓Teams maintaining an existing AutoGen codebase
- ✓Anyone who wants pattern vocabulary before picking a current agent framework
Prerequisites
- •Basic Python coding experience, as stated on the course page
- •An OpenAI-compatible API key if you run the notebooks outside the platform
- •Some idea of what an LLM tool call is and why agents need one
FAQ
What is AI Agentic Design Patterns with AutoGen?
A 1 hour 35 minute DeepLearning.AI short course taught by AutoGen's own creators, working through the four agentic design patterns — reflection, tool use, planning and multi-agent conversation — as six runnable projects: a stand-up comedy duo, customer onboarding, blog-post critique, conversational chess, financial analysis with code execution, and an automated stock report.
Is AI Agentic Design Patterns with AutoGen free?
AI Agentic Design Patterns with AutoGen is free to access.
What level is AI Agentic Design Patterns with AutoGen for?
AI Agentic Design Patterns with AutoGen is aimed at a intermediate audience. Recommended background: Basic Python coding experience, as stated on the course page, An OpenAI-compatible API key if you run the notebooks outside the platform, Some idea of what an LLM tool call is and why agents need one.
How long does AI Agentic Design Patterns with AutoGen take?
Expect roughly 1 hr 35 min, self-paced — 8 video lessons, 6 code notebooks, 1 graded quiz. Most learners work through it at their own pace.
What will I learn from AI Agentic Design Patterns with AutoGen?
You'll learn: Implement reflection, where a reviewer agent critiques and improves another agent's draft; Build two-agent and sequential-chat conversations that carry context forward; Give agents tools and constrain their actions to legal moves; Attach a code executor so agents can run real analysis themselves; Coordinate a planner agent with a group of specialist agents; Use nested chats and summarisation between stages of a conversation; Judge which agentic pattern actually fits a given task.
Topics
Sources
This page was written from 2 sources, 1 on domains other than deeplearning.ai.