Compose a multi-agent system in LangGraph without losing control of state.

In Lesson 3.8 of GROW Course 3 (Production Agentic Systems), Doug Liles works through it with you. A multi-agent pattern in LangGraph that stays debuggable.

What you'll take away

- When multi-agent earns its cost

- Graph and nodes

- Shared state

Chapters

(0:00) Cold open & the promise

(2:52) When multi-agent earns its cost

(5:44) Graph and nodes

(8:37) Shared state

(11:29) Control and handoffs

(14:22) Debugging the graph

(17:14) The artifact & what to do Monday

Study guide, slide deck, blog companion, infographic (free)

- https://fde.goodcombinator.ai/lessons/FDE_Lesson_3_8

Audio companion is AI-narrated using NotebookLM. Explainer video is human-produced.

Public release

This full FDE Certification audio lesson is available publicly as part of GROW by Good Combinator.

Continue in GROW Skills Library: https://goodcombinator.com/grow/library

Episode metadata

Show: GROW by Good Combinator

Course: Course 3 - Production Agentic Systems

Lesson: 3.8

Season/Episode: 3/8

Runtime: 20:07

Access label: public-full-lesson

Source channel: spotify

Source audio: app/lessons/assets/course-3/3.8/3.8 audio podcast.m4a

Taxonomy: Domain=Production Agentic Systems; Career stage=FDE; Skill level=Intermediate; Asset type=Lesson; Outcome=Build

CRM events: course_started, lesson_completed, cert_enrolled

Transcript and accessibility

Final transcript required before launch: https://goodcombinator.com/grow/courses/course-3/lesson-3.8/transcript

Accessibility requirement: reviewed audio, clear title, descriptive show notes, and transcript availability before launch.

Audio companion is AI-narrated using NotebookLM and disclosed here.

Credits

Host: Doug Liles

Producer: Good Combinator

Audio: NotebookLM AI narration (disclosed)