Anthropic gives away thirteen courses through Anthropic Academy, and AI Central's document lists every one of them as free. Nine of the thirteen state their size, and those nine come to 309 lectures and 29.4 hours of video. The catalog splits into two halves, short orientation courses of roughly an hour, and eight-hour platform certifications for developers deploying Claude through the API, Amazon Bedrock or Google Cloud's Vertex AI. Pick the half you actually need.
Reviewed 7 August 2026. Every figure below comes from AI Central's document, 13 Free Courses From Anthropic, which captures each Anthropic Academy course card as it appeared when the document was made.
What is actually on the list
Thirteen courses, and they are not thirteen versions of the same thing. Sorted by what they teach, the list falls into five groups.
- Five AI Fluency courses: AI Fluency Framework and Foundations, AI Fluency for Educators, AI Fluency for Students, Teaching AI Fluency, and AI Fluency for Nonprofits.
- Three platform courses: Building with the Claude API, Claude with Amazon Bedrock, and Claude with Google Cloud's Vertex AI.
- Two on Model Context Protocol, one introductory and one advanced.
- Two on Claude Code: Claude Code in Action, and Introduction to Agent Skills.
- One general primer, Claude 101, described as covering everyday work tasks and core features.
That distribution is the finding. Almost half of Anthropic's free catalog is not about writing code. It is about teaching students, faculty and nonprofit staff how to work with an AI system, which is a curious place for a model lab to spend its teaching budget.
It is two catalogs wearing one badge
Nine of the thirteen cards state a length. Added up, those nine come to 309 lectures and 29.4 hours of video. Three courses account for nearly all of it.
Building with the Claude API is 84 lectures and 8.1 hours. Claude with Amazon Bedrock is 88 lectures and 8 hours. Claude with Google Cloud's Vertex AI is 65 lectures and 8 hours. Those three alone are 237 of the 309 lectures, and 24.1 of the 29.4 hours.
Take them out and the remaining six counted courses total 72 lectures and 5.3 hours between them. AI Fluency for Students is 5 lectures and half an hour. Teaching AI Fluency is 7 lectures and 0.6 hours. Claude Code in Action is 15 lectures and one hour.
So there is no middle tier. You either give up most of a working day to a platform certification, or you spend a lunch break on an orientation course. Nothing here asks for three hours. Treat the list as one syllabus and you will either drown or feel short-changed, which is why sequencing matters more than signing up.
How these numbers were counted
Straight addition of the lecture, video and quiz figures printed on the course cards inside the document. Nine cards carry them. Four do not, namely Claude 101, AI Fluency for Educators, AI Fluency for Nonprofits and Introduction to Agent Skills, so those four are excluded from every total here. The real catalog is therefore larger than 309 lectures and 29.4 hours, never smaller. Eight of the nine counted courses also list quizzes, 36 in all, and those same nine cards each promise a certificate of completion.
AI Fluency is the largest single block
Five of the thirteen courses carry AI Fluency in the title, aimed at four separate audiences, plus one course on how to teach and assess the framework itself. The foundation course is described on its card in a single line.
Learn to collaborate with AI systems effectively, efficiently, ethically, and safely
That is Anthropic Academy's own wording, reproduced in AI Central's document. Read the four adverbs. Effectively and efficiently are the ordinary promises. Ethically and safely are the tell. This is a curriculum built for institutions that have to defend an AI policy to a board, a faculty senate or a donor, which is a real gap in the free training market.
The nonprofit edition makes the same bargain explicit.
This course empowers nonprofit professionals to develop AI fluency in order to increase organizational impact and efficiency while staying true to their mission and values.
Again, the card's own wording as captured in AI Central's document. Efficiency and mission in one sentence. If you run a team where people are quietly nervous about AI, that framing survives contact with the room, and it costs nothing.
MCP and Agent Skills are where the developer track is heading
Three of the thirteen cover plumbing that did not exist a couple of years ago. Two are Model Context Protocol courses, and one covers Agent Skills inside Claude Code. The introduction is pitched at connection, not theory.
Learn how to build modular AI applications using MCP to connect Claude with external tools and data sources
The document also captures the course goals printed on that page: what problems MCP solves, the split between clients and servers, tools the server exposes through a JSON schema, resources exposed as data, and prompts that guide model behaviour. Sixteen lectures, one hour, built around MCP servers in Python.
The advanced course adds sampling, notifications and transport implementations, in 15 lectures, 1.1 hours and two quizzes. Barely two hours takes you from not knowing what MCP is to understanding its transport layer, which is the best value on the list by a distance.
Introduction to Agent Skills is the newest idea here. The document describes Skills as reusable markdown instructions Claude applies automatically to the right tasks at the right time, and the course runs from building a first Skill to distributing them across teams.
What to do with this
The mistake is to open the list and start at number one. Start from what you are trying to ship.
- New to Claude entirely. Claude 101 for everyday work tasks, then AI Fluency Framework and Foundations, 14 lectures and 1.1 hours.
- Developer calling the models directly. Building with the Claude API, then Claude Code in Action, then Introduction to Model Context Protocol.
- Already on AWS or Google Cloud. Skip the generic API course and take the matching one, Bedrock at 88 lectures or Vertex AI at 65.
- Teaching or training people. Teaching AI Fluency at 7 lectures and 0.6 hours, then AI Fluency for Educators for institutional strategy.
- Running a nonprofit or small team. AI Fluency for Nonprofits for the framing, then Claude 101 for the people who will use the tool.
One warning worth acting on. All three platform courses share identical wording, each described as covering the full spectrum of working with Anthropic models, differing only in whether that runs through the Claude API, Amazon Bedrock or Vertex AI. Taking two is sixteen hours learning the same thing twice. The Bedrock course is additionally described as part of the accreditation program created for AWS, which matters if your employer wants a formal credential.
AI Central closes the document on a line that is really the whole point.
Don't wait for inspiration
That closing note sits under a page headed you've got the tools, now put them to work. Free courses get bookmarked and never opened. Thirteen certificates sit there at zero cost, and the shortest takes thirty minutes.
Are Anthropic's courses actually free?
Yes. All thirteen cards captured in the document show the word free beside the enrol or register button. No price, no trial period and no paid tier appears on any of them.
How long do Anthropic's free courses take?
It depends entirely on which one. Of the nine that state a length, the shortest is AI Fluency for Students at 5 lectures and half an hour, and the longest is Building with the Claude API at 84 lectures and 8.1 hours. Those nine total 29.4 hours.
Do you get a certificate from Anthropic Academy?
Nine of the thirteen course cards state a certificate of completion, including all three platform courses and both Model Context Protocol courses. The remaining four cards do not mention certification either way.
Is there anything on the list for non-technical people?
Close to half. Six of the thirteen need no code at all, namely Claude 101 and the five AI Fluency courses aimed at educators, students, nonprofit professionals, instructors and general learners. The other seven sit on the developer track.