🎭 Multi-Agent Workflows

Welcome to the advanced tutorials! In this course, you'll learn how to combine multiple ECC agents into powerful collaborative workflows — like commanding an elite development special forces team.

Prerequisites

We recommend completing Hello ECC and TDD Masterclass before starting this tutorial.

🎮 Experience Multi-Agent Collaboration

In real-world development, a complex feature often requires multiple experts working together. Try these command combinations:

ECC Command Simulator

❯

Available Commands:

📚 Three Core Orchestration Patterns

Multi-agent orchestration comes in three core patterns. Let's master them step by step:

Master the Three Orchestration Patterns

Step 1 of 4

1

Pattern 1: Sequential Orchestration

Agents execute in a fixed order, with each agent's output feeding into the next. Like an assembly line — each station completes its task before passing the work along.

# Sequential Pipeline
/plan Implement user auth     # Step 1: Planner creates plan
/tdd --feature=auth          # Step 2: TDD Guide develops
/code-review                 # Step 3: Reviewer audits
/e2e Test auth flow          # Step 4: E2E Runner validates

💡Best for standard feature development. Each agent's output automatically becomes the next agent's context.

💻 Design Your Multi-Agent Pipeline

In the code playground below, try designing a multi-agent workflow configuration for a real feature:

Multi-Agent Pipeline Configurationyaml

# multi-agent-pipeline.yaml # Try editing this pipeline configuration! name: "Feature Development Pipeline" description: "Standard workflow for new feature development" stages: # Stage 1: Planning (Sequential) - name: planning agent: planner input: "feature requirement" output: "implementation plan" # Stage 2: Development (Sequential per module) - name: development agent: tdd-guide mode: sequential modules: - "data-model" - "business-logic" - "api-endpoints" - "ui-components" # Stage 3: Review (Parallel) - name: review mode: parallel agents: - code-reviewer # Code quality - security-reviewer # Security audit # Stage 4: Validation (Conditional) - name: validation condition: "review.passed == true" agent: e2e-runner fallback: agent: tdd-guide action: "fix issues and re-review" # Configuration config: max_retries: 3 timeout_per_stage: "30m" notify_on_failure: true

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

❓ Knowledge Check

❓

Which orchestration pattern is best suited for 'multiple experts reviewing the same code simultaneously'?

ASequential Orchestration

BParallel Orchestration

CConditional Orchestration

DRecursive Orchestration

❓

In sequential orchestration, what's the best approach when the TDD Guide agent's tests fail?

ASkip the tests and continue to the next agent

BGo directly to code review

CReturn to TDD Guide to fix before continuing

DTerminate the entire pipeline

❓

What is the most critical principle when designing multi-agent workflows?

AUse as many agents as possible

BUse parallel mode for all tasks

CDefine clear responsibility boundaries for each agent

DEach agent should only run once

🎉 Congratulations!

You've completed the Multi-Agent Workflows tutorial! You've mastered:

• ✅ Sequential Orchestration: Assembly-line development process
• ✅ Parallel Orchestration: Multi-dimensional simultaneous review
• ✅ Conditional Orchestration: Intelligent routing decisions
• ✅ Combined Orchestration: Real-world project workflow design

📖 Next Steps

• Custom Hooks Development - Learn to create automation hooks
• E2E Testing Strategies - Master end-to-end testing