A reading companion for the Anthropic Academy course catalog. Anthropic's 20+ official courses — from everyday Claude use to AI fluency, Claude Code, the API, and MCP — are regrouped here into five thematic parts with a “choose your track” guide, so a reader of any background can find the route that fits them.
Each chapter distils one course into its key concepts, paired with a quiz. These are independent study notes; take the free, hands-on courses on Anthropic Academy for the full experience, including videos and exercises.
Original work & credit
Anthropic
Free, official Claude training courses from Anthropic Academy.
- Title
- Claude Mastery
- Publisher
- Anthropic Academy
- Source
- Anthropic Academy courses
These notes are independent, original summaries created for study purposes. All rights to the underlying work belong to Anthropic and Anthropic Academy. Please support the course creators by taking the original training.
Chapters
26 of manyStart Here: Choose Your Track
How this shelf is organized into five parts, and seven persona tracks — newcomer, knowledge worker, student, educator, business lead, new developer, AI engineer — each with a recommended reading route.
AI Capabilities and Limitations
A working mental model of generative AI — the four properties (next-token prediction, knowledge, working memory, steerability) as capability–limitation continua, how training shapes behavior, and diagnosing failures.
AI Fluency: Framework & Foundations
The 4D framework for collaborating with AI — Delegation, Description, Discernment, Diligence — plus the three modes (automation, augmentation, agency), six prompting techniques, and the Description–Discernment loop.
Claude 101
Everyday Claude — what it is, writing effective prompts, the three shapes of desktop work (Chat, Cowork, Code), Projects, Artifacts, Skills, connectors, Enterprise Search, Research, and the wider product family.
AI Fluency for Students
The 4Ds for student life — using AI as a learning partner without outsourcing your thinking, applying it to career planning, and being the human in the loop with academic integrity.
AI Fluency for Educators
The 4Ds for faculty and instructional designers — applying AI Fluency to course design and learning outcomes, and to materials and assignments built for a world where students have AI too.
Teaching AI Fluency
Teaching and assessing the 4Ds — the two loops (Delegation–Diligence, Description–Discernment), assessing the competency not the artifact, and grounding it in your discipline.
AI Fluency for pK-12 Educators
A two-course path (with the American Federation of Teachers) — the 4Ds for the pK-12 classroom with a strong responsible-use emphasis, then a Train-the-Trainer kit to teach other educators.
AI Fluency for Small Businesses
The 4Ds for owners and small teams — the inner loop (refining work with AI) and outer loop (transparent, human-in-the-loop use) to grow impact while staying true to your mission.
AI Fluency for Nonprofits
The 4Ds for mission-driven teams (with GivingTuesday) — researching and writing with AI, privacy and data analysis, and workflow augmentation and integration.
AI Fluency for Creative Work
The 4Ds for creators — the creative-value and production lenses for where AI belongs, the inner/outer loops, and AI's roles as Delegation outcomes, all while protecting your voice.
AI Fluency for Builders
The 4Ds for people who build products — Delegation and the builder's toolkit, Discernment split into code and user experience, and Diligence: standing behind what you ship.
Introduction to Claude Cowork
Delegating whole tasks to Claude — Chat vs Cowork vs Code, the three patterns of work that fit, delegating end to end, standing context/skills/plugins, and working safely.
Claude Code 101
Agentic coding in your codebase — what Claude Code is, the explore → plan → code → commit workflow, CLAUDE.md for persistent memory, and the subagents/skills/MCP/hooks customization features.
Claude Code in Action
Long, less-supervised runs you can trust — steering (plan, compact, rewind, goal, loop, worktrees), configuring guardrails, automating repeat work, and verifying unsupervised runs by the diff.
Introduction to Agent Skills
Teach Claude once — SKILL.md and on-demand matching, where skills live, configuration and multi-file skills with progressive disclosure, and how skills compare to CLAUDE.md, subagents, hooks, and MCP.
Introduction to Subagents
Focused helpers with their own context — what subagents are and the built-ins, the four design patterns, and the decision rule: does the intermediate work matter?
Claude Platform 101
Building on Claude programmatically — the three platform layers, the Messages API, the agent loop and tool use, and handing work off with managed agents.
Building with the Claude API: Fundamentals
The API foundation — accessing the Messages API, multi-turn conversations and system prompts, and controlling output with temperature, streaming, and structured data.
The Claude API: Evaluation & Prompt Engineering
Measure before you tune — the prompt-evaluation workflow and test datasets, code-based vs model-based grading, and the core prompt-engineering techniques.
The Claude API: Tool Use, RAG & Features
From talking to doing — tool use and its loop, Retrieval-Augmented Generation for grounding, and Claude's features: thinking, vision, PDFs, citations, prompt caching, and code execution.
The Claude API: MCP, Apps & Agents
The last mile — the Model Context Protocol (clients, servers, tools/resources/prompts), Anthropic apps and computer use, and the workflow-vs-agent orchestration patterns.
Claude on Amazon Bedrock
Running Claude through AWS — why Bedrock, accessing Claude via the Bedrock client with IAM and model IDs, and the fact that all the API capabilities transfer unchanged.
Claude on Google Cloud
Running Claude through Vertex AI — enabling it in Model Garden, gcloud application-default credentials, and the same Claude toolkit (plus a few extra RAG and apps topics).
Introduction to Model Context Protocol
Building MCP servers and clients from scratch — why the protocol exists, the host/client/server architecture, and a server's three capabilities: tools, resources, and prompts.
MCP: Advanced Topics
Production MCP — sampling (a server borrowing the client's model), log/progress notifications, roots as a scope boundary, and the transports: JSON-RPC, STDIO, and StreamableHTTP.