Developers
$149once
The full developer track. Four modules, every template and skill file.
- Developers track — 4 modules, all files
- Lifetime access and every update
- 14-day refund, one email
Track 01Developers
Claude Code course for developers.Stop prompting. Start engineering.
Most developers use Claude like a faster autocomplete. This track teaches the operating model: skills that encode how your team works, agents that run unattended, and tool integrations that make Claude part of your stack instead of a chat window beside it.
Who it is for
- Engineers already using Claude Code who feel they only touch 20% of it
- Tech leads deciding how a team should adopt agents without chaos
- Solo builders who want to ship products faster with fewer regressions
- Anyone who has written a prompt file and wondered why it keeps drifting
By the end
You will be able to
- Write skills that Claude picks up reliably, with tests that prove it
- Design subagent workflows for review, research and refactors
- Wire MCP servers so Claude can act on your real systems safely
- Set up hooks, permissions and worktrees for hands-off runs
- Debug agent failures with a repeatable method instead of vibes
Developers
What you will work with
The same building blocks Anthropic documents for Claude Code, taught in the order they actually matter. Each one gets a working file you keep.
- 01
CLAUDE.md and memory
The operating manual Claude reads first. What belongs there, what belongs in the repo, and how to keep it from drifting.
- 02
Claude skills (SKILL.md)
How to create custom Claude skills that trigger when they should, with evals that prove it. The core of the course.
- 03
Subagents and agent workflows
Fan-out research, review gates, adversarial verification, and deterministic orchestration for unattended runs.
- 04
Hooks and permissions
Guardrails the model cannot skip: pre-commit checks, formatting, and permission modes for hands-off sessions.
- 05
MCP servers
Give Claude tools on your real systems. Design tool schemas an LLM can use, build a server, and debug what it actually sent.
- 06
Worktrees and plugins
Isolated workspaces for parallel agents, and packaging skills as plugins your whole team installs.
Curriculum
4 modules. Open any of them.
01Claude Code fundamentalsThe mental model. Context, permissions, memory and the loop you are actually running.
- What Claude Code is doing between your prompt and the diff
- CLAUDE.md as an operating manual, not a wish list
- Permission modes, and when auto mode is the right call
- Memory: what to persist, what to leave in the repo
- Session hygiene: worktrees, scratchpads, when to start fresh
- Lab: audit an existing CLAUDE.md and cut it by half
02Skills that stickWriting SKILL.md files Claude actually triggers, with evals that prove it.
- Anatomy of a skill: frontmatter, trigger description, procedure, references
- Why most skills never fire, and how to write descriptions that do
- Progressive disclosure: keep the skill short, link the depth
- Testing skills: trigger evals, outcome evals, regression sets
- Packaging skills as plugins for a team
- Lab: turn a wiki page your team ignores into a skill they use
03Agents, subagents and workflowsFan-out, review gates and unattended runs without losing the plot.
- When a subagent helps and when it just costs tokens
- Briefing an agent: goal, ruled-out paths, files to read, scope
- Review patterns: fresh-eyes review, adversarial verification, two-stage gates
- Deterministic orchestration with workflow scripts
- Hooks: pre-commit checks, formatting, guardrails the model cannot skip
- Lab: a code-review pipeline that runs on every branch
04MCP and tool integrationGive Claude hands on your real systems, with the blast radius you choose.
- MCP in one diagram: servers, tools, resources, prompts
- Designing tools an LLM can use well: names, schemas, error text
- Building a server in TypeScript or Python, from scratch to installed
- Auth, scopes and read-only defaults
- Debugging tool calls: what the model saw, what it sent, why it failed
- Lab: an MCP server for the internal API your team keeps re-explaining
Both tracksBest value
$199once
Everything in both tracks. For people who build and run the work.
- Developers track — 4 modules, all files
- Non-Developers track — 4 modules, all files
- Lifetime access and every update
- 14-day refund, one email
Questions
Questions developers ask
What is the difference between Claude skills, MCP and subagents?
Skills are instructions: how to do a task. MCP servers are tools: what Claude can reach, such as your database or an internal API. Subagents are workers: separate Claude sessions you dispatch for parallel or isolated work. Most real workflows use all three, and the Developers track teaches when each one is the right choice.
How do I create a custom Claude skill (SKILL.md)?
You write a SKILL.md file with a name, a description that tells Claude when to use it, and the procedure. The hard part is the description: most skills never trigger because it is vague. Module 2 of the Developers track covers this in depth, with evals that prove your skill fires when it should.
Do I need a Claude Pro or Max subscription to take this course?
The Developers track uses Claude Code, which needs a Claude Pro or Max plan or API access. The Non-Developers track works on any paid claude.ai plan, including Pro. The course price does not include a Claude subscription.
How long does it take to learn Claude Code or finish a track?
Each track is four modules of two to three hours. Most people finish a track in two weeks working an evening or two a week, and are using their first skill on real work after module two.
How quickly will this Claude course go out of date?
Claude changes monthly, so the material is revised as it does and you get every update for free, forever. Modules are written around the operating model, which changes far more slowly than the feature list.
All questions on the pricing page. Something else? Email Ethan directly.