What is Claude Academy?

What it actually is

Claude Academy is Anthropic's own education site for learning to work with Claude and to build on it. Anthropic frames it around three audiences in one line - people exploring AI, people getting started with Claude, and people rolling it out to a team - and that framing is the most useful thing on the site, because it tells you which third of the library is yours.

Three practical facts before anything else. It is free. Browsing needs no account - you can read without signing in at all, and signing in with a Claude account is what saves your progress. And it is one library, not a linear program: there is a Get started entry point and an All resources view, and nothing in between tells you where to stop.

The All resources view reports 292 resources, filtered by All, Courses, Tutorials, and Use cases.- counted on academy.claude.com/all, 1 September 2026

Every number on this page carries that date. Anthropic adds to the Academy continuously, so treat counts, lesson totals, and hour estimates here as a reading taken on one day rather than a specification. The shapes and the organising logic move much more slowly, which is why this article leads with those.

If what you want is the course list itself - every course with its lesson count, hours, and badge - that is a separate page: Learn Claude. This one is the map of the building; that one is the reading list.

By the numbers

The whole library in six figures, read on one day. The first three describe how it is organised; the last three describe how much of it is course material with real weight - the part you can actually finish.

Two of these are the ones to hold onto. 292 is why nobody should read the Academy front to back - it is a library, and the shapes in the next section are how you find the shelf. 54 hours is the ceiling, not the plan: one track for one exam is a quarter of that, and the Claude prep plan does that arithmetic per credential.

Counted on academy.claude.com on 1 September 2026. Resources from the All resources view; courses, hours, and badges from the 17 main courses in the catalog, excluding the eight short AI Fluency audience variants.

Four shapes of resource

This is the part worth internalising. The Academy's Resources menu offers Courses, Tutorials, Use cases, and Webinars, and they are not four labels for the same thing - they differ in length, in whether anyone assesses you, and in what you walk away holding. Most frustration with the Academy is a shape mismatch: reaching for a 9-hour course when you needed a 7-minute tutorial, or skimming a use case when you actually needed the assessed course.

One wrinkle worth knowing, because it will confuse you otherwise: the All resources filter exposes only three of the four - courses, tutorials, and use cases - while webinars sit in the Resources menu alongside them. So "all resources" means all self-paced resources. If you are hunting for a session, go through the menu, not the filter.

The practical routing rule: tutorial to unblock, use case to copy, course to actually learn, webinar to hear it argued. Every resource in the library carries its shape and its weight on the card, so the cost of a wrong turn is visible before you click - which is more than most vendor academies offer.

The five product areas

Above the shapes, the Academy is organised by product, and the five areas are the five ways Anthropic expects you to meet Claude. Anthropic's own one-line framing for each is the fastest way to work out which is yours.

• Claude.ai. Thinking through problems, drafting documents, and analysing data in conversation. The everyday-work surface, and where most people should start.
• Claude Cowork. Handing off whole tasks and getting polished deliverables back, for yourself and your team - delegation rather than conversation.
• Claude Code. Building, debugging, and shipping in your terminal, IDE, and browser.
• Claude Tag. Tagging Claude into your team's channels and handing it work everyone can follow - the collaborative surface, and the one most readers do not know exists.
• Claude Platform. Building Claude into your products with the API, the Claude Console, and MCP.

The split that matters is between the first four and the last one. Claude.ai, Cowork, Code, and Tag are about using Claude at increasing degrees of delegation; the Platform area is about building with it, and it reads like developer documentation rather than training - which is the right call, and the subject of section 06.

AI Fluency and the 4D framework

Sitting beside the product areas is a collection with no product attached: AI Fluency. It is the part of the Academy that outlasts any model release, and Anthropic gives it top-level billing rather than filing it under a product - product features age, reasoning habits do not.

The spine is the 4D framework: Delegation, Description, Discernment, and Diligence - decide what to hand over, describe it well, judge what comes back, and stay responsible for it. Three courses carry the core:

• AI Fluency: Framework & Foundations. The 4D framework itself, taught for effective, efficient, ethical, and safe collaboration. 14 lessons, 1 quiz, 4 hr, completion badge.
• AI Capabilities and Limitations. An accurate mental model of what language models can and cannot do - next-token prediction, knowledge, working memory, steerability, and context limits. 13 lessons, 1 quiz, 3.5 hr.
• Teaching AI Fluency. For faculty and instructional designers who have to teach and assess it, not just practise it. 7 lessons, 1 quiz, 4.5 hr.

The same framework is then re-cut for specific audiences in shorter courses - builders, creative work, educators, pK-12 educators, nonprofits, small businesses, and students. They are the same four ideas applied to different daily work, so take exactly one: the one whose examples look like your week. Taking two is repetition, not depth.

The developer track

If you write code against Claude, the Platform area is the Academy's densest material by a wide margin - and the one place where the library stops being introductory. It also links outward to Anthropic's API reference, cookbooks, and help centre rather than trying to restate them, which is the correct division of labour.

The order that works, shortest first:

• Claude Platform 101. Written for people who have made a few API calls or only ever used the chat window. 13 lessons, 1 quiz, 1.5 hr.
• Building with the Claude API. The heavyweight: prompting, tool use, RAG, agents, MCP, and production patterns. 67 lessons, 8 quizzes, 9 hr. Plan it across two weeks, not an evening.
• Introduction to Model Context Protocol. MCP servers and clients from scratch with the Python SDK, across the three core primitives - tools, resources, and prompts. 10 lessons, 1 quiz, 1 hr.
• Model Context Protocol: Advanced Topics. Sampling, notifications, and roots, with walkthroughs of each protocol flow. 11 lessons, 1 quiz, 1.5 hr.

If you reach Claude through a cloud provider rather than directly, two courses mirror the API one almost lesson for lesson: Claude with Amazon Bedrock (65 lessons, 8 quizzes, 8 hr) and Claude with Google Cloud's Vertex AI (66 lessons, 9 quizzes, 8.5 hr). Take the one matching your deployment - they overlap heavily, so taking two is mostly rereading.

Around the edges sit the agent-layer shorts: agent skills (6 lessons, 1 hr), subagents (4 lessons, 45 min), and Claude Code in Action (9 lessons, 1 quiz, 1 hr) for long hands-off sessions you can actually trust. Cheap, and they are where the day-to-day leverage is.

Rolling it out to a team

The third audience in Anthropic's framing - rolling Claude out to a team - gets one substantial course, and it is the least discussed thing in the Academy: Deploying Claude Enterprise with Confidence. 14 lessons, 1 quiz, 2.5 hr.

Its structure is the useful part even if you never take it. The course is built around five decisions that shape a rollout - Structure and Identity, Access, Governance, Spend, and Visibility - and it ends with a rollout plan you keep rather than a badge you post. If you are the person who has to answer procurement, that list is a better meeting agenda than anything you would write from scratch.

Read it next to Claude Team vs Claude Platform, which covers the same territory from the billing and org-shape side.

What it is not

One correction worth making plainly, because the naming invites the mistake: the Academy is not the certification program. Free courses and paid proctored exams are two different things run by the same company.

• The Academy is free and open. No purchase, no company email, no eligibility check. Sign in only if you want progress saved and badges earned.
• The exams are paid and gated. Four credentials, $99 to $175 per attempt, and registration expects a company email tied to a Claude Partner Network organization. Full detail on Claude certification.
• Neither requires the other. No course is a prerequisite for any exam, and Anthropic's exam guides treat recommended experience as advisory - the credential is awarded on exam performance alone.

A completion badge from a course is not a certification, and saying otherwise on a CV is the kind of thing an interviewer checks. It is still worth having: it says you finished something structured, which is more than most people can show.

For most readers the honest advice is to stop at the Academy. Take one AI Fluency course and one product course, use Claude on real work while you do it, and sit an exam only when a client, an employer, or a partner listing actually asks for one. If you want the reading list rather than the map, go to Learn Claude next.