Build reliable multi-agent workflows for complex software development
Why Attend?
Single-agent coding workflows can break down as tasks become longer, more complex, and more parallel.
The core problems we usually run into are:
- Context degradation
- Self-grading
- Sequential execution
- Hidden specification drift
- Unbounded retries
This workshop shows you how to solve those problems by designing engineering graphs with clear node boundaries, structured handoffs, parallel execution, independent verification, and controlled recovery.
The goal is simple: move from one agent doing everything to a workflow where each agent has a defined job, clear boundaries, and an independent way to prove the work is correct.
NOTE: In case you aren't able to purchase a ticket from Eventbrite, you can do so from Luma. Here's the link: https://luma.com/packt-4p1d
What you’ll be able to do
By the end of the workshop, you’ll be able to:
- Recognize when a single-agent workflow is no longer enough
- Decide when a graph is justified and when it is not
- Break development work into specialized, parallelizable nodes
- Define clear completion and failure criteria for each node
- Control agent boundaries through tools, models, and worktree isolation
- Pass structured artifacts instead of bloated conversation history
- Add independent verification and deterministic quality gates
- Detect specification drift even when tests pass
- Design bounded retry, remediation, and escalation paths
- Balance reliability against cost and complexity
Who should attend?
This workshop is designed for:
- Developers already using Claude Code
- Senior and Staff Engineers
- AI Engineers
- Technical Leads and Architect
- Platform and Developer Productivity Engineers
- Teams exploring subagents, agent teams, or multi-agent coding workflows
What you’ll receive
- Three-hour live, instructor-led workshop
- Guided multi-agent workflow build
- Practical frameworks and implementation patterns
- Reusable graph-engineering mental model
- Complete workshop recording
- Workshop slide deck
- Live Q&A with the instructor
- Certificate of completion
Stop asking one agent to do everything
Move beyond isolated Claude Code sessions and learn how to engineer a reliable, coordinated workflow for planning, implementation, testing, review, and recovery.
Reserve your place now.
Good to know
Highlights
- 3 hours
- Online
Refund Policy
Location
Online event
Agenda
-
Keynote by Ken Huang
-
Where Single-Agent Workflows Break
A feature built by one long-running agent loop The defect it shipped while reporting the work as complete Why a passing test suite is not the same as correct work Workshop structure and the repository you received
-
From Agent Loops to Agent Graphs
The core agent loop: discover, plan, execute, verify Trigger, task, and verification: the three parts of every loop Why a graph is several connected loops rather than a new pattern Nodes, edges, state, and gates Knowledge graphs and agent graphs: the distinction