Anthropic Dropped 13 Free Courses — I Broke Down Every Single One

#ai #productivity #claude #beginners

Anthropic quietly put 13 courses on a platform called Skilljar. All free. Certificates included. You don’t even need an Anthropic account — just sign up with your email and you’re in.

But nobody has time for 13 courses. I certainly didn’t.

So I went through every curriculum, scraped every public summary, review, and GitHub note I could find, and distilled the whole thing.

The verdict: only 5 matter if you’re a developer. The rest are for educators, students, and nonprofits.

The 5 You Should Take

Claude Code in Action

21 lessons, about an hour. If you use Claude Code in your terminal, this is priority number one.

It covers persisting memory across sessions with CLAUDE.md, turning repetitive workflows into slash commands, auto-formatting files with Hooks, connecting GitHub and databases via MCP servers, and embedding Claude Code into your own tools with the SDK.

One course, and your Claude Code game completely changes.

Building with the Claude API

16 lessons, the most comprehensive one.

It starts from API basics and goes all the way through prompt evaluation, Tool Use, RAG pipelines, Extended Thinking, Prompt Caching, MCP, and Agent/Workflow patterns.

If you’re building any kind of AI-powered service, this is the bible.

Introduction to Agent Skills

A newer course.

It teaches you how to create, configure, and distribute reusable Skills (markdown-based instructions) in Claude Code. Build a Skill once, and Claude automatically applies it to the right tasks. Covers team distribution and troubleshooting too.

Introduction to Model Context Protocol

8 lessons.

You build MCP servers and clients from scratch in Python. The three core primitives — Tools, Resources, Prompts — and testing with the MCP Inspector.

MCP is the standard for connecting AI models to external services. Understanding it properly opens up a lot of possibilities.

Model Context Protocol: Advanced Topics

8 lessons.

The sequel. Sampling (where the server asks the AI model for help), Notifications (async updates), and Transport mechanisms (stdio, SSE, Streamable HTTP) for production deployments.

What About the Other 8?

Claude 101 covers basic Claude features. If you already use Claude daily, there’s not much new.

AI Fluency: Framework & Foundations teaches a thinking framework called the 4D model. It’s more theoretical — worth skimming, not deep-diving.

Claude with Amazon Bedrock and Claude with Google Vertex AI are each 16 lessons, but they overlap about 70% with Building with the Claude API. The difference is that every example uses AWS SDK (boto3) or GCP SDK respectively. Take these when you’re actually deploying to a specific cloud.

AI Fluency for Educators, for Students, for Nonprofits, and Teaching AI Fluency — these four target teachers, students, and nonprofit professionals. They were co-developed with professors from Ringling College and University College Cork, released under Creative Commons.

Great for educational institutions. Not relevant for developers.

My Recommended Order

Week one: finish Claude Code in Action and Introduction to Agent Skills. Both are short.

Week two: take Building with the Claude API. This one’s dense — budget two to three days.

Week three: do the MCP intro and advanced courses back to back.

Five courses, and you’ve covered the core of the Claude ecosystem. Total time is under 10 hours.

One more thing — beyond Skilljar, Anthropic has free Jupyter Notebook courses on GitHub: API Fundamentals, Prompt Engineering Tutorial, Real World Prompting, Prompt Evaluations, and Tool Use.

If you prefer running code over watching videos, start there.