Anthropic Academy: All 7 Free Claude Courses, Ranked

Anthropic runs a whole academy of free courses on Claude, and I did the seven core ones cover to cover. It took about 14 hours. What I found is that the order almost everyone takes them in is wrong, that one of them is the single best free AI course I have come across anywhere, and that two of them you can genuinely skip depending on who you are.

So here is the honest ranking, the exact path for a beginner, a developer and a security person, and a straight answer on whether the certificates are worth anything, which turns out to be a more interesting question than it sounds. Everything below comes from doing the courses and from reading Anthropic's own course pages, not from summarising other people's summaries.

I Ranked All 7 Free Anthropic Claude Courses (Which One Actually Wins)

The video walks through all seven out loud with the courses on screen. This article goes further: the exact lecture counts, video hours and stated prerequisites for each one, the three learning paths, and the certificate question the video does not get into. Read it, watch it, or both.

What is Anthropic Academy?

Anthropic Academy is Anthropic's official training platform for Claude. It is hosted on Skilljar rather than on anthropic.com, and every course on it is free. Anthropic's own FAQ is explicit that you do not need an Anthropic account to take them: the courses only require a Skilljar account, which is an email and a password, with no payment step.

As of today the public catalog holds 20 individual courses plus one learning path, and the path contains two more courses inside it. When I recorded the video there were around 13. They keep adding niche ones, mostly on the AI Fluency side, so the number you read in an older article is almost certainly out of date.

The seven I am ranking here are the core set: the ones that teach Claude itself and the technical stack around it, rather than the audience-specific AI Fluency spin-offs for educators, students, nonprofits and small businesses.

The 7 core courses, ranked

A note on the hours before you plan your week. Only four of the seven publish a video duration, and those four add up to 11.3 hours of video. Claude 101, Claude Code in Action and Introduction to agent skills publish no duration at all, so any total you see quoted for the full set is somebody's estimate, including mine. My 14 hours is real time spent, exercises included, not a figure off a page.

1. AI Fluency: Framework & Foundations

This is the winner, and it surprised me, because it is not about Claude's technology at all. It is about how you actually think and work alongside an AI, and it is packed with exercises, so you are doing rather than reading.

The spine is the 4D Framework: Delegation, Description, Discernment and Diligence. Delegation is deciding what to hand over in the first place. Description is telling it clearly. Discernment is judging what came back. Diligence is owning what you ship. That last section on diligence is better than anything I have read on the subject anywhere else, and it is the one people skip because it sounds like a compliance module. It is not. It is the part that stops you shipping something you did not check.

Here is the detail almost nobody mentions and it explains why the course is so much better than the rest. It was not written by a marketing team. The framework was developed in 2023 and 2024 by two academics, Prof Rick Dakan of Ringling College of Art and Design and Prof Joseph Feller of University College Cork, and the courses were produced by Dakan, Feller and Anthropic with support from the Higher Education Authority in Ireland. The materials are released under a CC BY-NC-SA licence, so any institution can adapt them. On the Anthropic listing it is taught alongside Drew Bent and Maggie Vo of Anthropic. It reads like a university course because it is one.

Time: Anthropic's page lists 1.1 hours of video and a final assessment. The framework's own site is more honest about the real cost, calling it three to four hours including roughly 70 minutes of video. The gap is the exercises, and the exercises are the point.

My take: take it, whoever you are. It is the only course in the set that makes you better with every AI tool you touch rather than one, and months later it is the one still paying you back.

One more thing worth knowing: the course carries an AI diligence statement disclosing that Claude 3.7 helped develop it, with the human authors making all final decisions. A course about diligence that discloses its own use of AI is a small thing that tells you a lot about the people who made it.

2. Building with the Claude API

The big one. 84 lectures, 8.1 hours of video and 10 quizzes, which makes it roughly as large as the other six put together. It is the most complete technical course in the set and it is clearer than the same topics scattered across the others.

It runs through seven sections: getting started with Claude, prompt engineering and evaluation, tool use, retrieval augmented generation, Model Context Protocol, Claude Code and computer use, and finally agents and workflows, with a final assessment at the end. Prompt caching, extended thinking, streaming, structured output and multimodal handling of images and PDFs are all in there.

Prerequisites, stated on the page: proficiency in Python, basic knowledge of handling JSON, and your own Anthropic API key. The course is free. The API calls you make while doing it are not, so budget a few dollars.

Two honest shortcuts. If you have already done AI Fluency, skip the prompting section here, because you will have covered the thinking already and this version is the applied one. And if you can read code, skip watching someone read it out loud and just work through the exercises.

My take: absolutely worth it if you are building with AI. If you are not, it is eight hours you will not get back.

3. Introduction to Model Context Protocol

Have you ever tried to understand MCP and just bounced straight off? Same. I watched tutorials, I read the docs, and it never clicked. This is the course that finally made it click, and that alone earns it third place.

It covers what MCP is, how the pieces fit together, and how to stand up a server, a client, resources and prompts, with files you follow along with. You build a document management system and test it with the MCP Inspector, which is the moment the whole thing stops being an abstraction. 16 lectures, one hour of video, one final assessment.

Prerequisites, stated on the page: basic Python, an understanding of async and await patterns, and familiarity with API concepts. They tell you up front, which I appreciated.

My take: if you are building with AI at all, take it. MCP is fast becoming the way every AI system plugs into everything else, and being the person in the room who actually understands it is worth an hour.

4. Claude Code in Action

This is the practical one for people who already write code. It is not an introduction to Claude Code, and the page says so: the prerequisites are that you already use Claude Code for single prompts and have basic familiarity with Git and the command line. This course is about what comes after that.

Nine lessons across four themes: steering long sessions, configuring Claude with a CLAUDE.md that actually gets followed, verification skills, permission modes and hooks, then automating repeat work with routines, headless runs, GitHub Actions and code review, and finally verifying unsupervised runs and using plugins. There is a course quiz at the end.

One honest flag: the stats block on the course page reports zero video hours and zero quizzes while the curriculum clearly lists a quiz, so the metadata is stale. Do not plan around a duration for this one.

My take: essential if you are a developer. Skip it entirely if you never plan to code. And if you are a developer who has not yet coded alongside an AI, this is the course that closes that gap. If you are not a developer but the idea appeals, start with our own plain-English walkthrough instead: Claude Code Simplified builds a real working page without a line of code.

5. Claude 101

The guided tour. What Claude is, your first conversation, getting better results, working on your desktop, projects, artifacts, skills, connectors, enterprise search, research, and a closing section on use cases by role. Thirteen lessons plus a certificate quiz, no prerequisites, no code.

It is a clean way to get your bearings if you just signed up, instead of clicking around lost. But be honest with yourself: if you already use Claude every day, you will pick up most of this in an afternoon of playing. A fair amount of what it demonstrates is also on paid plans, so on the free tier some of it can feel like a shop window.

My take: take it if you are brand new. Skip it if you are not. Either way, bookmark the skills and connectors lessons, because those two are the genuinely handy part and they are the ones people come back for. We cover the same ground from a different angle in 6 Claude Features Most People Don't Use.

6. Introduction to agent skills

Short and focused. Six lessons: what skills are, creating your first one, configuration and multi-file skills, how skills differ from other Claude Code features, sharing them, and troubleshooting them. That last lesson is the genuinely useful one, because troubleshooting is where skills actually fall over.

It is the lightest course in the set. There is no quiz and no assessment item in the curriculum, so it is videos and reading rather than practice. Nothing to install, nothing to build, no language to learn, because a skill is just Markdown with some frontmatter.

My take: short enough that anyone can watch it, and skills are one of the highest-leverage things in the whole Claude ecosystem, so the payoff per minute is good. Just do not expect anything hands-on.

7. Model Context Protocol: Advanced Topics

This goes deep. Sampling, log and progress notifications, roots, JSON message types, the STDIO transport, the StreamableHTTP transport and how state works across it. 15 lectures, 1.1 hours of video, two quizzes, and it pairs each concept with a written walkthrough. The hands-on modules are the best in the whole set.

I will be straight with you: it is dense, and I nearly nodded off in places. Not because it is badly made, the opposite, but because none of it sticks unless you are actively building something that needs it. Reading about transport layers you have never had a problem with is exactly as memorable as it sounds.

Prerequisites, stated on the page: a basic understanding of MCP servers and clients, and familiarity with async programming patterns.

My take: last place is not a criticism of the course, it is a criticism of taking it too early. Do the introduction, go and build something real, hit an actual problem, then come back. It will be a different course when you do.

The right order for you

The ranking is not the path. The path depends on who you are, and this is where most people waste hours.

If you are new to AI or Claude and you do not code:

If you are a developer or a builder:

If you work in security and you are sizing up the risk:

What about the other 15 courses?

The seven above are the core. The rest of the catalog splits into three groups.

More Claude product courses: Claude Code 101, Claude Platform 101, Introduction to Claude Cowork, Introduction to subagents, and AI Capabilities and Limitations. If Claude 101 felt too thin, Claude Platform 101 and the Cowork course are where to go next.

AI Fluency for a specific audience: AI Fluency for educators, for students, for nonprofits, for Small Businesses, and for Builders, plus Teaching AI Fluency and a pK-12 educators learning path with two courses inside it. These are the same framework aimed at different rooms. If one of those labels is you, take that version instead of the general one, not as well as it.

Cloud platform integration: Claude with Amazon Bedrock and Claude on Google Cloud. Take these only if your company already runs on that cloud.

Do the free certificates actually count?

This is where I would slow down, because there are two different things wearing the same word.

The free Academy courses issue completion certificates. Claude 101 has a "Certificate of completion" as an explicit item in its curriculum, and the AI Fluency page states that you take a final assessment and receive a certificate of completion. Several of the others do not name a certificate anywhere in their curriculum, and Introduction to agent skills has no assessment at all, so treat "all seven give you a certificate" as unproven. These are attendance records. They say you finished a free course, which is a real and honest thing to say, and it is not a credential.

Anthropic's actual certification program is separate, and it is paid. It runs on a different Skilljar site for partners, and the published exam prices are $99 for Claude Certified Associate, $125 for Developer Foundations, $125 for Architect Foundations and $175 for Architect Professional. Those come with a Credly badge that Anthropic explicitly says you can share on LinkedIn. The free course certificates carry no such published statement, so if someone tells you a free Academy certificate is a LinkedIn credential, they are blending the two programs together.

And one genuinely counter-intuitive detail. AI Fluency: Framework & Foundations is also on Coursera, free to enrol, listed at three hours, with around 5,100 people enrolled and a 4.7 rating from about 100 reviews. But the Coursera page states plainly that no certificates, credentials or reports are awarded there. So the free version on Anthropic's own site gives you the certificate and the Coursera version does not. If the certificate matters to you, take it on Anthropic Academy. If you just want the teaching and you prefer Coursera's player, either is the same course, and the underlying materials are CC licensed anyway.

What none of the seven teach you

Here is the honest note to end on, and it is not a criticism, because these courses are genuinely good and mostly free of the fluff that fills this category.

They will make you fluent in the tool. Not one of them tells you what to build with it, or how to turn any of it into money. You can finish all seven, hold every certificate, understand transports and sampling and prompt caching, and still be sitting in front of a blank screen with no idea which thing to make. Fluency is not the same as ownership.

That gap is the whole reason IdeasRepay exists. We take the fluency courses like these give you and turn it into something you own.

Come and answer the question the courses do not

The nearest starting point is free. Our Build Your AI Team course walks you through building your first AI employee, the Writer, with the prompt pack included, and our free Claude prompting course drills the one skill every course above quietly assumes you already have.

When you want the business rather than the skill, that lives at ideasrepay.com. Every walkthrough there is something you click through and build, not something you read: the validation you do before writing a line, the tool stack with current prices, the prompts in the order you run them, the templates and scripts as real downloads, what to charge, and a first-week checklist that ends at a paying customer.

The closest one to what you just learned is The Vibe-Coded Micro-SaaS. It takes the exact skills these courses gave you, the API, the prompting, the agent work, and points them at a small software product with your name on it: how to prove people want it before you build it, how to build it with AI, how to price it, and where the first people who pay actually come from.

One payment of $99 opens that walkthrough and every other one we have published, plus every one we publish after it, across all three verticals: online businesses, YouTube and content, and offline work in the real world. Downloads and templates included, no renewals, no upsells, and you can email us while you build. That is launch pricing, and it moves to $199 after the first 500 members. There is a community building alongside you and I mentor through it personally.

Go get fluent with these. Then come and build something of your own.