Anthropic Academy: Free Claude Courses, and Which to Take

Anthropic Academy is Anthropic’s free catalog of self-paced Claude courses, from AI basics for anyone to multi-hour developer courses on the Claude API, Claude Code and MCP. As of September 30, 2026, it is moving to a new home and a new name: Claude Academy, at academy.claude.com. The old address, anthropic.com/learn, now redirects there, and the original Anthropic Academy site on Skilljar still works but points courses that have moved to their Claude Academy copies. Each course is a set of short lessons, most end in a quiz, and finishing one earns a completion badge. Which to take depends on what you do: the paths below cover everyday users, developers, builders on the API and admins.

Anthropic Academy, Claude Academy and the Partner Academy

Three names turn up in search, and they are not the same thing:

• Claude Academy↗ (opens in a new tab) is the current home. It describes itself as “Free courses, tutorials, and use cases from Anthropic’s education team,” and its catalog mixes courses with shorter tutorials and use cases. You sign in to save progress.
• Anthropic Academy is the earlier name, on the Skilljar course site↗ (opens in a new tab). Its pages still carry the Anthropic Academy name, and a banner on moved courses says the course “now lives on Claude Academy.” People who take courses through an employer can keep using it.
• The Anthropic Partner Academy is a separate site for companies in the Claude Partner Network. It holds the prep courses and registration for Claude certification exams, which the free academies do not.

The courses, grouped by who they suit

This is the course list on Claude Academy’s all resources page↗ (opens in a new tab), with the lesson counts and times it shows. Times are Anthropic’s estimates.

For anyone who uses Claude at work

• Claude 101 (13 lessons, about 2.5 hours): from your first conversation and prompting to projects, artifacts, skills and connected tools.
• AI Fluency: Framework and foundations (14 lessons, about 4 hours): Anthropic’s 4D framework of Delegation, Description, Discernment and Diligence for working with AI well.
• AI capabilities and limitations (13 lessons, about 3.5 hours): what large language models can and cannot do, including context limits.
• Introduction to Claude Cowork (14 lessons, about 2.5 hours): handing multi-step work on your own files and projects to Claude in Cowork.
• Introduction to Claude Tag (11 lessons, about 2.5 hours): Claude in your team’s Slack workspace. Claude Tag itself is in beta.
• Building effective human-agent teams (5 lessons, about 45 minutes, marked beta): preparing a team to work with AI together rather than one person at a time.

For developers using Claude Code

• Claude Code 101 (12 lessons, about 1.5 hours): what Claude Code is, how it works and the core workflows.
• Claude Code in Action↗ (opens in a new tab) (9 lessons, about 1 hour): for people already using Claude Code on single prompts who want longer, less supervised sessions. It covers plan mode and rewind, CLAUDE.md instructions, permission modes and hooks, routines and GitHub Actions, and checking unattended runs.
• Introduction to agent skills (6 lessons, about 1 hour): building, configuring and sharing skills in Claude Code.
• Introduction to subagents (4 lessons, about 45 minutes): splitting work across parallel subagents.

For developers building on the API

• Claude Platform 101 (13 lessons, about 1.5 hours): building on the Claude Platform from the ground up, even if you have only used Claude in a chat window.
• Building with the Claude API↗ (opens in a new tab) (67 lessons, 8 quizzes, about 9 hours): the long one. Modules cover API access, prompt evaluation, prompt engineering, tool use, RAG and agentic search, MCP, Claude Code and computer use, agents and workflows, and Claude’s features. It expects Python, basic JSON and an API key, and API calls are billed to your own Console account.
• Introduction to Model Context Protocol↗ (opens in a new tab) (10 lessons, about 1 hour): building MCP servers and clients in Python around tools, resources and prompts. The Skilljar site also lists a follow-on, Model Context Protocol: Advanced Topics, covering sampling, notifications and transports.
• Claude with Amazon Bedrock and Claude with Google Cloud’s Vertex AI (about 8 and 8.5 hours): the API course’s material for teams that reach Claude through those clouds.

For admins, educators and specific fields

• Deploying Claude Enterprise with confidence (14 lessons, about 2.5 hours): five rollout decisions, structure and identity, access, governance, spend and visibility, recorded in a rollout plan.
• The AI Fluency family applies the 4D framework to a setting: builders, creative work, educators, pK–12 educators (made with Teach For America, plus a train-the-trainer kit), nonprofits, small businesses and students.

Which course to take first

1. You use Claude for writing, research or analysis: Claude 101, then AI Fluency: Framework and foundations. Add Introduction to Claude Cowork if you want Claude to work through files for you.
2. You are a developer new to Claude Code: Claude Code 101, then Claude Code in Action. Take agent skills and subagents once you repeat the same instructions often.
3. You are adding Claude to a product: Claude Platform 101 for orientation, then Building with the Claude API in chunks, then Introduction to Model Context Protocol if your app will call outside tools.
4. You are rolling Claude out to a company: Deploying Claude Enterprise with confidence, plus Claude 101 for the people you roll it out to.
5. You teach or lead a team that is new to AI: the AI Fluency course for your setting, which is less about buttons and more about judgment.

Courses teach the moves; the habits that make them pay off come from use. For the Claude Code habits that matter most day to day, such as planning first, giving Claude a check it can run and clearing context between tasks, see Claude Code best practices rather than waiting for a course to cover them.

Completion badges are not certification

A free course ends in a quiz and a completion badge. That shows you finished it; it is not a Claude certification. Anthropic’s certifications, such as Claude Certified Architect – Foundations, are proctored exams delivered by Pearson VUE, currently open only to people at companies in the Claude Partner Network, with their own prep courses in the Partner Academy. What the architect exam covers and how to prepare for it is in Claude Certified Architect: what the exam covers. The public API, Claude Code and MCP courses above are a sound base for it.

Turning a course into a plan you finish

Nine hours of video is easier to finish as small pieces you can see moving. Paste a course’s module list into a board as tasks and move each one from To Do through In Progress to Completed as you go. In fenbs that is Add many, which turns a pasted list into tasks in one step; each task’s note can hold what you want to try from that module, and its plan the small project you will build to practice. If you study with a colleague or an AI assistant, they can see the same board and comment on it. fenbs has no due dates, so the pace is yours to keep, and Copy as Markdown gives you the whole list back as text when you want to share progress.