About This Course
<p>In this hands-on course, you will move beyond AI-assisted coding and learn to lead Claude Code through the design and development of a real distributed microservices system, working in the terminal, one spec at a time and one feature branch at a time.</p><p>You will use Spec-Driven Development (SDD) to turn requirements into written specs. Claude Code plans each spec in Plan Mode, implements it on its own Git feature branch, verifies it, and merges it into main. The capstone EShop is built with C#, .NET 10 and Aspire, but the workflow itself does not depend on the language: the same CLAUDE md, WOW md, specs, skills and subagents work in a Java, Python, Node.js or Go repository.</p><p>This is not a beginner prompt-engineering course. You act as the AI Architect: you write the specs, set the rules and review every change, while Claude Code explores the codebase, plans the work, writes the code and commits it.</p><p>Inside this course, we will focus on:</p><p>Claude Code Agent Harness and Architecture: The agentic loop (gather context, take action, verify results), the built-in tools, sessions, the context window, checkpoints and the four permission modes: Manual, Accept Edits, Plan and Auto. Understand why Claude Code behaves the way it does, so you can steer it reliably.</p><p>The Explore → Plan → Code → Commit Workflow: Anthropic’s recommended loop for every change. Claude explores the code, proposes a plan you approve in Plan Mode, implements it, verifies it with builds, tests and running services, and finishes with a descriptive commit.</p><p>Planning the Project with Claude: PROJECT md and PLAN md: Before any code exists, Claude interviews you with the AskUserQuestion tool. Your answers become a PROJECT md with the scope, architecture and constraints, and a phased PLAN md. The first phase becomes a running distributed skeleton.</p><p>Spec-Driven Development for Microservices: Learn the SDD levels (spec-first, spec-anchored, spec-as-source) and compare GitHub Spec Kit, Kiro and OpenSpec. Then build your own lightweight SDD workflow on Claude Code features: a WOW md (Way of Working) rulebook for humans and agents, a spec template, and a spec backlog that Claude breaks out for each microservice.</p><p>Feature Branch Management with Claude Code: One rule drives delivery: one spec = one sub-feature = one feature branch. The spec is the first commit on its branch, the code is implemented and verified there, and the branch is merged into main with git merge --no-ff, so every change stays small and reviewable.</p><p>Develop Distributed Microservices with Claude Code: Build the Catalog, Basket and Ordering services, each owning its own data in PostgreSQL or Redis, plus a Blazor storefront, all orchestrated with Aspire. Prove the services talk with a trace in the Aspire dashboard, and let Claude test the buyer’s journey in a real browser with Claude in Chrome.</p><p>CLAUDE md and Context Engineering: Bootstrap project memory with /init, then write an effective CLAUDE md with your architecture rules, conventions and Git workflow. Measure and control context with /context, @ references, /compact and /clear, resume sessions, and undo mistakes with /rewind checkpoints.</p><p>Agent Skills for the SDD Loop: Turn what you repeated into four delivery skills. /new-spec writes or improves a spec, /start-spec runs preflight checks, creates the branch and plans the work, /finish-spec develops, verifies and merges, and /auto-spec chains all three in one command. Then build the Ordering microservice with them.</p><p>Subagents for Review and Testing: Create two specialists, each with its own context window. The spec-compliance-reviewer audits every branch against its spec before the merge, and the integration-smoke-tester walks the end-to-end buying journey across Catalog, Basket and Ordering. Then build the Basket microservice with skills and the reviewer together.</p><p>Each part of the EShop adds one new capability: Catalog with the manual SDD loop, Ordering with skills, Basket with subagents. You learn each feature at the moment the previous step shows you need it.</p><p>If you are a software engineer, tech lead or solutions architect ready to go from using an AI assistant to leading an agentic development workflow, enroll today and start building distributed systems with Claude Code.</p>
What you'll learn:
- Lead Claude Code as the AI Architect: you write the specs and review every change, while Claude explores, plans, codes and commits
- Understand how Claude Code works inside: the agentic loop, built-in tools, context window, sessions, checkpoints and the four permission modes
- Use Spec-Driven Development (SDD) to turn requirements into written specs, and make each spec the source of truth for Claude Code’s work
- Apply the Explore → Plan → Code → Commit workflow to every change, approving Claude’s plan in Plan Mode before any code is written
- Deliver every change with one rule: one spec = one sub-feature = one feature branch, verified and merged into main with git merge --no-ff
- Write an effective CLAUDE md with /init, and manage context with /context, @ references, /compact, /clear and /rewind checkpoints
- Build Agent Skills that automate the whole spec lifecycle: /new-spec, /start-spec, /finish-spec and /auto-spec
- Create custom subagents: a spec-compliance-reviewer that audits each branch against its spec, and an integration-smoke-tester across services
- Build Catalog, Basket and Ordering microservices with PostgreSQL, Redis and a Blazor storefront, orchestrated and observed with Aspire