Plan the fix across three files, then edit in small batches.
scope locked · three files · one prompt · diff resolving under review
CLAUDE CODE COURSE
Learn Claude Code on real tickets, step by step
This Claude Code course teaches agentic work the way it happens on the job — resolving in real time, plan first, edit across files, verify with tests. Watch a diff rewrite itself, then learn why.
Work from ticket-style tasks in a sample repo. Evenings in Cambridge and live online.
FIT CHECK
Who picks up this track quickly
Developers comfortable with the command line and version control settle in fast. You should be able to clone a repo, run its tests, and read a diff. Career switchers with that baseline do well if they complete the prep notes first.
If the terminal still feels new, start with the Cursor AI course and return here second. That path keeps the agentic loop intact without fighting setup. Our AI coding training meets you where your tooling already works.
Learn AI-powered development across Claude Code, Cursor AI and OpenAI Codex — this track focuses on the repo-wide loop.
CASE FILE — THE LOOP
Agentic loop you will repeat
Every module follows the same loop: read the ticket, inspect the relevant files, ask for a plan, approve a bounded edit, run checks. Instructors show how to keep long edits from drifting by freezing scope in writing. You practice stopping the assistant mid-run when the plan slips.
- read ticket → inspect files → approve plan
- bounded edit → run tests → narrow retry
- repetition is deliberate; the loop is the skill
small batches beat large rewrites — the bench exercise mirrors the site idea
CASE FILE — PLANNING + EDITS
Planning before generation
You will write short plans that name files, expected behavior, and tests to run. Claude Code then works against that plan rather than improvising. We show how much context is enough: linked modules yes, whole repo no. You keep a one-page planning template for later use at work.
Multi-file edits without losing track
Real tickets touch several files at once. You practice reviewing each file row in the diff, running the test tied to that change, and requesting a narrower retry when one file goes wrong.
WORKING HABITS
Debugging with the assistant
You will feed a failing test and the surrounding function, ask for two candidate causes, then verify by running code rather than trusting the explanation. Instructors point out confident misdiagnoses and how to catch them. Each debug ends with a regression note in the commit message. That note matters more than the fix.
Guardrails teams actually keep
We teach secret handling, dependency caution, and mandatory human review before merge. You will practice redacting examples and checking lockfile changes line by line. The habits line up with governance-aware workflows discussed around tools like Amazon Q Developer in enterprise settings. Bring your team review policy and we will map it to the loop.
REVIEW RECORD
TICKET PRACTICE AND REVIEW SESSIONS
Half of class time is ticket practice in pairs with instructor walkthroughs. You present your diff, explain what you accepted and rejected, and hear how another pair handled the same ticket.
Review language stays concrete: what changed, what test proves it, what remains risky. No grades, just marked checklists you can reuse on Monday.
SCHEDULE, FORMAT, AND PREP
Evenings in Cambridge, same exercises online
Evening cohorts meet at 77 Massachusetts Ave, Cambridge, MA 02139, with the same ticket exercises taught live online. Plan on a laptop with a working dev environment; setup instructions arrive with the syllabus.
Prep takes about three hours: terminal basics, test commands, and a short reading on agentic coding. These AI coding courses stay practical from the first session.
Not sure about readiness? Call +1-617-253-1000, Monday to Friday, 9:00 AM – 6:00 PM, or write to [email protected]. We will walk through your setup and point you to the right starting track.
prep: ±3 hours · format: evenings + live online
QUESTIONS
Before you pick this course
Short answers that help you decide between Claude Code training and our editor-led tracks.
How is Claude Code different from an editor assistant?
It works well from a prompt across a repo rather than beside your cursor. You describe the ticket, it proposes and edits multiple files, and you verify. Editor tools stay closer to the line you are typing.
Do I need advanced prompt skills first?
No. We teach goal, context, and constraints together inside each exercise, so prompt habits grow with tool habits. By the end you can run a careful loop of scope, prompt, review, and commit.
Where do I continue after this track?
Many learners continue with the Agentic Workflows Guide, then compare the Cursor AI course or the OpenAI Codex course to round out their AI coding training. Contact us for a syllabus and format comparison.