Multi-Agent Pipeline Training¶
A structured training curriculum for building and operating multi-agent AI pipelines in Cursor. Learn concepts, practice hands-on, then design and architect.
The pipeline¶
A multi-agent pipeline is a sequence of specialized agents that pass work forward via JSON artifacts. Each agent has a clear role; artifacts create traceable handoffs.
graph LR
Planner -->|delegates| Subplanner
Subplanner -->|plan.json| Worker
Worker -->|worker-result.json| Tester
Tester -->|test-result.json| Reviewer
Reviewer -->|review-result.json| Git
Tester -->|FAIL| Debugger
Debugger -->|debug-diagnosis.json| Worker
Git -->|git-result.json| PR
When tests fail, the Debugger diagnoses the issue and the Worker retries. The Planner orchestrates the full flow.
Three tiers¶
-
Foundation
Can you understand and use AI effectively?
6 exercises — concepts and vocabulary
-
Practitioner
Can you build and deploy AI features?
11 exercises — hands-on pipeline use
-
Expert
Can you architect AI systems and lead others?
8 exercises — tiered routing and cost optimization
Who is this for?¶
Teams adopting AI-assisted development with Cursor or Claude Code. This curriculum teaches you how to set up multi-agent pipelines, delegate work to specialized agents, and scale from learning to production.
- Getting Started — adoption guide
- Quickstart — first pipeline in 5 minutes
- Reference Cheat Sheet — single-page quick reference