Anthropic teaches AI for free. 21 courses and a month plan

Every week someone asks me which AI course they should buy. Usually it is a program costing a few hundred dollars, half of which retells the documentation. My answer is almost always the same: start with the free material, and get it straight from the people who build the models.

Anthropic, the company behind Claude, runs a learning academy. The learning section of its site points to academy.claude.com, and the full catalog lives on the Skilljar learning platform. I went through all of it and put together a route that fits into a month of evenings.

What to know before you enroll#

• The courses are free: the registration button on course pages is marked free.
• You do not need an Anthropic account, only an account on the learning platform itself. To put what you learn into practice, you will need access to Claude.
• The catalog holds 21 courses, from a first introduction to AI all the way to building agents.

For everyone, even if you have never written code#

AI Fluency: Framework and Foundations. The backbone of the academy: about four hours, 14 lessons and a quiz. It teaches you to work with AI effectively, efficiently, ethically and safely using Anthropic's four-part framework. If you take only one course, take this one.

AI Capabilities and Limitations. Three and a half hours and 13 lessons on how a model really works: next-token prediction, where its knowledge comes from, how working memory behaves and where the context limits are. Afterwards you stop expecting miracles where there will be none and start getting results where they are possible.

Claude 101 and Introduction to Claude Cowork. A hands-on introduction to the products: how to set tasks and work with an assistant in everyday work.

Industry versions of AI Fluency. Separate courses for small businesses, creative work, nonprofits, educators and students, plus one for people who want to teach AI fluency themselves.

Building Effective Human Agent Teams. A short 45-minute beta course on moving from working with a single assistant to a team of people and agents. For managers these may be the most useful 45 minutes in the entire academy.

For people who build products#

Claude Code 101 and Claude Code in Action. The first introduces the tool, the second is aimed at developers who already use Claude Code for one-off prompts. It has 10 lessons in four blocks: steering long sessions, configuring Claude for a project, automating repeat work and verifying runs that happen without supervision.

Claude Platform 101 and Building with the Claude API. Everything about working with the model through the API: requests, tools and how applications are put together.

Introduction to Model Context Protocol and MCP: Advanced Topics. MCP is the open standard that connects models to your data and services. If you build agents, you can no longer do without it.

Introduction to agent skills and Introduction to subagents. How to package repeatable skills for an agent and how to split a large task across several subagents.

AI Fluency for Builders, and for teams working in the cloud, dedicated courses on Claude in Amazon Bedrock and Google Cloud.

A one-month route#

Week 1. Foundations. AI Fluency: Framework and Foundations plus AI Capabilities and Limitations, about eight hours in total. Assignment: pick three work tasks that repeat every week and hand them to AI using the new rules.

Week 2. Tools. Claude 101, Claude Cowork and the AI Fluency course for your field. Assignment: build a personal library of five tested prompts and templates.

Week 3. Automation. If you write code, Claude Code 101 and Claude Code in Action. If not, Building Effective Human Agent Teams and another round of week two practice. Assignment: automate one routine task end to end.

Week 4. Agents. Introduction to Model Context Protocol, agent skills and subagents. Assignment: connect an agent to one real data source, such as a spreadsheet, your inbox or a knowledge base.

How to get more out of free than out of paid#

• Learn on your own tasks. A course gives you the frame, but the skill only shows up when you practice on your own material.
• Show results in public. A completion badge says you watched the lessons. A working automation, a repository or a case write-up says you can actually do it. I wrote about this in detail in the piece on how AI ate the first rung of the career ladder.
• Count the money early. If the courses lead you to build on the API, read my breakdown of how to stop overpaying for AI.
• Do not tie yourself to one provider. The principles in these courses carry over to any model.

The best knowledge about AI is free today. The only price is your time, and that is the one investment no model can make for you.