Claude Code course guide: official training, curriculum, and certification

Claude Code course decision summary

Fast official start

Claude Code 101 is listed with 12 lectures and about one hour of video.

Broader official path

Claude Code in Action is listed with 15 lectures, about one hour of video, and one quiz.

Focused official modules

Introduction to subagents is listed at 20 minutes, while Introduction to agent skills is listed at 30 minutes.

Integration depth

Introduction to Model Context Protocol is listed with 16 lectures, about one hour of video, and one quiz.

Live-demo route

Anthropic's June 2026 foundations workshop covers installation, a first fix, repository rules, skills, hooks, subagents, MCP, plugins, and product surfaces.

Certification boundary

a course completion record, third-party certificate, and Anthropic technical certification are different credentials.

Stronger evidence

a reviewed project can show planning, diffs, tests, security decisions, and recovery beyond passive video completion.

Which official Claude Code course should come first?

Claude's current course directory groups structured learning by product and category. The public cards show lecture count, video length, and quiz count when listed. The course itself opens through Anthropic's Skilljar learning site, where registration and curriculum details can differ by module.

Official 101 course

12 lectures and about one hour of video in the current Claude course directory.

The in-action course

15 lectures, about one hour of video, and one quiz in the current directory.

Subagents module

four lectures and about 20 minutes of video.

Agent skills module

six lectures and about 30 minutes of video.

MCP introduction

16 lectures, about one hour of video, and one quiz.

Related depth

the same directory also lists Claude Platform, API, Bedrock, Vertex AI, AI fluency, and advanced MCP courses.

Review the current Anthropic course catalog before enrolling.

What does Claude Code in Action cover?

The current registration page describes comprehensive software-development training and lists free registration. It expects terminal familiarity and basic Git knowledge. Its published curriculum moves from assistant architecture and setup through context, changes, automation, MCP, GitHub, hooks, the SDK, and a final quiz.

Foundations

what a coding assistant is and how Claude Code acts on a repository.

Setup

installation, project preparation, adding context, and making changes.

Context control

managing relevant information and referencing project resources.

Automation

reusable commands or workflows, hooks, and SDK concepts.

Connections

MCP servers and GitHub integration for external tools and review workflows.

Planning

choosing thinking and planning approaches for work with different complexity.

Course language can lag product terminology. For example, current Claude Code documentation treats custom commands as part of skills. Treat the course as concepts and practice. Verify command names, settings, permissions, and product behavior in the current documentation.

Open the current Claude Code in Action registration page and curriculum.

How can the official workshop support practice?

Anthropic's Claude Code Workshop: Foundations was a one-hour event on 11 June 2026. Its page offers a recording route and describes a live path from cold installation to a fixed bug. The curriculum also covers repository rules, Plan mode, skills, hooks, subagents, MCP connectors, plugins, and supported surfaces.

• Watch once to understand the product loop and current vocabulary.
• Repeat the first-fix exercise inside a disposable repository rather than following passively.
• Pause before every action and predict which context, permission, tool, and validation should apply.
• Compare the recorded interface with the current version because product surfaces change quickly.

Open Anthropic's Claude Code Workshop: Foundations page.

Build a project-based Claude Code curriculum

A course becomes useful when each lesson produces reviewed evidence. Use one small repository with known behavior and tests. Learners should explain the task, predict risk, supervise the change, inspect the diff, run independent checks, and recover from one intentional mistake.

Module 1: install and set a safe boundary

• Install from the official path and record the version, authentication route, provider, and update method.
• Open a disposable repository with no secrets or production access.
• Identify the project root, instruction files, ignored content, test commands, and protected paths.
• Use default or Plan mode and inspect every requested permission.
• Stop a running action, reject a proposed edit, and restore the original state.

Module 2: clarify the goal and context

• Write one goal with reader or user outcome, scope, constraints, acceptance checks, and excluded work.
• Reference the smallest useful files, lines, diagnostics, logs, issue, design, or documentation.
• Ask Claude to restate assumptions, unknowns, plan, risks, and validation before editing.
• Compare a vague prompt with a bounded prompt and record the difference in investigation and output.
• Add durable repository facts to CLAUDE.md and move repeatable procedures into skills.

Module 3: make and review one change

• Choose a small bug or feature with a failing test and a clear expected result.
• Inspect the plan before implementation and question one unstated assumption.
• Inspect every file change and command instead of accepting the final summary.
• Run the failing test, relevant suite, type check, lint, build, and a manual scenario.
• Ask for a self-review, then verify the result independently and document remaining uncertainty.

Module 4: extend only for a repeated need

• Create one skill for a repeated repository workflow and test its trigger on a difficult case.
• Add one hook for deterministic enforcement, such as blocking a protected path or running a targeted check.
• Connect one read-only MCP server and inspect its tools, data, authentication, permissions, and context cost.
• Use a subagent for isolated research or review, then audit what the summary omitted.
• Package extensions only when another project or team needs the same reviewed setup.

Module 5: operate the team workflow

• Define who can install plugins, connect services, change permissions, approve code, and release work.
• Use worktrees or branches to isolate concurrent tasks and prevent agents from overwriting shared files.
• Record usage, failed tasks, review time, reverted changes, escaped defects, incidents, and developer feedback.
• Review repository instructions, skills, hooks, MCP servers, accounts, and access on a named schedule.
• Retest the curriculum after major Claude Code, model, IDE, provider, or policy changes.

Teach goal clarity and discernment at every step

Anthropic's current research-backed curriculum identifies goal clarification as the signature starting behavior for Claude Code. It also says discernment does not grow automatically with product tenure. The suggested sequence is to teach the signature move, advance through ways of describing work, and revisit discernment at every step.

Goal clarity

state what should change, why it matters, what must remain, and how success will be checked.

Basic description

use prompt clarity, file references, and a deliberate model choice.

Middle description

shape repeatable work with skills, commands, and MCP connections.

Lasting description

use CLAUDE.md, hooks, and subagents to affect work across sessions.

Discernment

question assumptions, test difficult cases, inspect configuration, and verify evidence independently.

Read Anthropic's research-backed curriculum for Claude Code learning.

Use Better Design's installation guide for the first hands-on lesson.

Is there an official Claude Code certification?

Anthropic announced Claude Certified Architect, Foundations on 12 March 2026 as its first technical certification. The announcement says it is a partner-available exam for solution architects building production applications with Claude. It is broader than a Claude Code course and is not named Claude Code Certification.

Official product course

training listed in Anthropic's course directory and delivered through its learning platform.

Course completion record

evidence that a learner finished a provider's module or assessment, under that provider's rules.

Third-party certificate

a credential issued by a university, platform, professional chapter, instructor, or training company.

Anthropic technical certification

a formal exam named and announced by Anthropic, with the audience and access stated by Anthropic.

Portfolio evidence

reviewed work showing that the learner can scope, supervise, test, explain, and recover an agentic coding task.

Read Anthropic's announcement of Claude Certified Architect, Foundations.

When is a third-party course worth paying for?

Pay for a course when it adds something the official catalog and independent practice do not provide. Useful additions include instructor feedback, office hours, graded projects, and team facilitation. A domain-specific repository, accessibility review, security feedback, or a valued credential may also justify the cost.

• Coursera currently lists Claude Code: Software Engineering with Generative AI Agents as a beginner course with six modules and about five hours of work.
• The Coursera page describes a platform certificate path. Treat that as a Coursera course credential, not Anthropic's technical certification.
• IEEE Toronto advertised a four-hour June 2026 workshop with an Introduction to Claude Code certificate issued by IEEE Toronto.
• Treat each date, cohort, price, recording, instructor, support promise, assessment, refund, and certificate claim as provider-specific.

See the current Coursera Claude Code course details and certificate terms.

Verify the issuer and current status of the IEEE Toronto workshop certificate.

Use a Claude Code course scorecard

Freshness

publication date, last update, demonstrated Claude Code version, and current documentation links.

Prerequisites

coding, terminal, Git, repository, account, plan, API, editor, cloud, and operating-system needs.

Practice

real repository tasks, starter state, failing tests, review checkpoints, mistakes, and recovery.

Coverage

goals, context, tools, permissions, diffs, tests, instructions, skills, hooks, MCP, subagents, Git, and governance.

Assessment

what learners must produce, who reviews it, which rubric applies, and whether identity is verified.

Credential

exact issuer, title, requirements, expiration, verification method, transfer value, and renewal rules.

Support

instructor access, feedback time, community quality, office hours, captions, transcripts, and accessibility.

Cost

tuition, subscription, API usage, lab infrastructure, exam, retake, renewal, and learner time.

What should an evidence portfolio contain?

Ask every learner to submit one bounded project. Include its brief, repository state, plan, prompts, permissions, diff, tests, manual checks, review comments, rollback, and reflection. Remove secrets and private data. The artifact should show what the learner decided, what Claude executed, and how correctness was established.

Scope evidence

clear goal, excluded work, constraints, risks, and acceptance checks.

Context evidence

selected sources, repository instructions, assumptions, and unresolved questions.

Execution evidence

plan changes, tool use, permissions, rejected proposals, and human interventions.

Quality evidence

tests, static checks, manual scenarios, accessibility checks, and independent review.

Recovery evidence

one reverted mistake, failed test, or changed assumption with a documented response.

Learning evidence

what the learner would automate, restrict, or teach differently next time.

A practical learning path

Take Claude Code 101 for the short orientation, then complete Claude Code in Action beside a disposable repository. Add the focused skills, subagents, and MCP modules only when the project needs them. Connect the pieces with the official workshop, and consult current documentation during every exercise.

Choose paid training for feedback, assessment, team practice, or a valued issuer. Choose certification for the role it actually validates. In both cases, require project evidence that the learner can clarify, supervise, test, question, and recover Claude Code work.

Frequently asked questions