Two-day live virtual course
Your developers already have AI coding tools. Most are using them as autocomplete.
Claude Code for Developers is two days that take a team from AI-assisted typing to directing, governing and shipping agent-built software. Every developer leaves with a working product they built and deployed themselves.
Claude Code is Anthropic's coding agent.
- Inc. 500 firm
- 25k+ practitioners trained
- Scrum Alliance training partner
The problem
Accepting a suggestion and directing an agent are different skills. Only one of them gets taught, which is why most teams stop at autocomplete and never see the rest of the modes.
Four modes. Most teams only use one.
Developer-Driven, AI-Assisted
The developer writes. The agent accelerates.
AI-Driven Development
The developer directs and reviews. The agent writes.
AI Governance
The agent reviews every change before it merges.
AI in the Product
The model ships inside what your customers use.
Day one covers modes one and two. Day two covers three and four.
What they bring back to your codebase
A deployed, working product, that they can take away with them.
Project instruction files that hold the agent to your conventions on every session.
A pull request review gate they wired themselves and tested with a defect they planted.
Tests written from acceptance criteria before any implementation exists, which is how a claim of done stays honest when nobody typed the code.
How it runs
- Two days, delivered live and virtual.
- Mornings teach. Afternoons build.
- Private cohorts scheduled on request.
What they build
Everyone builds the same product, alone: a shared family calendar for one household. They choose their own stack with the agent, deploy it to a live address on day one, and keep it. The backlog is written by a Product Owner who is not in the room and is deliberately larger than two days, so the thing they carry out of Friday afternoon has somewhere obvious to go on Monday.
Who to send
Working developers who build web applications and are comfortable with git and a terminal. No prior AI tooling experience required. Rooms that mix sceptics with early adopters do well.
What each developer sets up beforehand, about an hourLearning objectives
Claude Code ships changes weekly. These objectives are revised between cohorts to match the current tooling.
See all 32 objectivesHide the 32 objectives
Day one · Modes 1 and 2
Why we are doing this
- What a coding agent is, and why it is not autocomplete
- The four modes, and where you sit today
Your first session
- What the agent can and cannot see
- The project brief, and why a generated one is a draft
- When it goes wrong, and what recovery looks like
Context and memory
- What context is, and why output gets worse
- Managing context on purpose
- Rules and working agreements as context that survives
Trust: tests, permissions, letting go
- Why tests matter more with an agent than without one
- Permissions and plan mode as trust controls
- Directing a build without touching the keyboard
Sprint 1 and review
- Choosing a stack with the agent, and why the decision is context
- What done means when someone else wrote the criteria
Sprint 2 and review
- Reviewing what you did not write, at the speed it arrives
- What the backlog did not say, and why that is normal
Day one close
- Turning today's friction into tomorrow's working agreement
Day two · Modes 3 and 4
What you wrote yesterday
- From a rule that helps you to a rule that holds when nobody is looking
The gate
- What an automated reviewer can and cannot be trusted to catch
- A gate is a contract with yourself, not a gatekeeper
- Why you break your own code on purpose
Secrets and things that fail
- What never enters the repository, and how you prove it
- When a third party fails, and how you would know
AI inside the product
- Building with AI is not the same as shipping AI
- Propose, wait, confirm
- What a model call costs, and what it does when it fails
Sprint 3 and review
- Building the guardrail before the thing it guards
- Demonstrating the refusal, not only the yes
Sprint 4 and review
- Sharing something with people who did not build it
- What you would tell the Product Owner now
Close
- Keeping current with a tool that changes weekly
- When the codebase is too big to fit
- Committing to the next three
Instructors
Fred Mastropasqua
Founder, CEO
Certified Scrum Trainer at ClearlyAgentic, with 30+ years of IT experience spanning enterprise software development, infrastructure, agile coaching, and executive transformation.
Robert Shaw
Certified SAFe Trainer, Agile Coach
An Enterprise Agile Coach and SAFe Trainer at ClearlyAgentic who specializes in helping large organizations move from Agile theory to enterprise-scale execution, coaching leaders, launching ARTs, and building the structures that make agility stick.
Questions
Do developers need experience with AI coding tools?
No. The first morning assumes none, and rooms that mix sceptics with early adopters do better than rooms that don’t.
Is this training for one specific tool?
The tool is Claude Code. The practices — context management, review gates, testing work you did not write, shipping a model call safely — transfer to whatever your team uses next.
What does each developer need set up beforehand?
About an hour of installs and accounts, listed on the prerequisites page. Two of them need budget approval, so we send the list two weeks out.
Prerequisites and course detailsDo they keep what they build?
Yes. It is deployed to their own account, on their own repository, and it keeps running after the course ends.
Is there a cost beyond the course fee?
Each developer needs a small credit balance on their own Anthropic API key for the day-two work, a few dollars. Everything else runs on free tiers.
Can it be delivered on site?
Virtual only for now. Private cohorts of this course are available on request.
Start the conversation
Claude and Claude Code are trademarks of Anthropic. This is independent training and is not affiliated with, endorsed by, or certified by Anthropic. Course details and registration.