Anthropic Documentation: 12 Pages to Bookmark (2026)

What it is: A curated list of the 12 most-useful pages in Anthropic’s official documentation for anyone learning Claude in 2026 — what each page covers, why it matters, and the order to read them in.
Who it is for: Anyone using or building with Claude who wants a source-of-truth bookmark stack rather than a thousand scattered tutorials. Especially useful for new Claude.ai Pro and API users.
Best if: You want the official docs ranked by usefulness, not by Anthropic’s internal site map.
Skip if: You’re looking for community tutorials or step-by-step videos — this catalog is Anthropic-official only. For daily AI news in one email, subscribe to our free daily newsletter.

Related deep-dive: For the Computer Use tool specifically, we’ve summarized Anthropic’s May 13, 2026 production best-practices post at Claude Computer Use Best Practices — covering screenshot resolution, model selection, prompt-injection defense, and context management.

Bottom line up front: Anthropic publishes hundreds of documentation pages at docs.anthropic.com, and most users miss the 12 that matter most. This post is the shortcut: the dozen pages that, if you read them in order, give you working competence with Claude faster than any third-party course can. Pages 1-5 are for everyone. Pages 6-12 are for developers and power users.

Learn Our Proven AI Frameworks

Beginners in AI created 6 branded frameworks to help you master AI: STACK for prompting, BUILD for business, ADAPT for learning, THINK for decisions, CRAFT for content, and CRON for automation.

Why bookmark Anthropic’s docs instead of relying on tutorials?

Third-party AI tutorials go stale fast. The Anthropic docs are the source of truth and they are updated alongside every model release, feature launch, and pricing change. The whole site is open, searchable, and surprisingly readable for an official documentation product. The catch: there are hundreds of pages and the site map prioritizes completeness over usefulness. Most beginners spend an hour wandering before realizing they should have just read three specific pages first.

This list cuts through. We use these 12 pages ourselves. Anchor your Claude knowledge to the official source and you stop worrying about whether a third-party tutorial is current.

The 12 pages, in order

1. Get started with Claude

URL: docs.anthropic.com/en/docs/get-started

The first page anyone should read. Covers what Claude is, the surfaces it runs on (Claude.ai, API, Claude Code), and the three-minute first request. Even if you’ve used Claude for months, this page is worth a re-read whenever a new generation ships.

2. Models — Opus, Sonnet, Haiku and their tradeoffs

URL: docs.anthropic.com/en/docs/about-claude/models

The current model lineup, context windows, capabilities, and the tradeoff matrix. Opus for hardest reasoning, Sonnet for the productive default, Haiku for high-volume cheap tasks. The model-name version numbers change — the doc keeps the canonical list. See our comprehensive AI models overview for the cross-lab picture.

3. Prompt engineering overview

URL: docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/overview

The most underrated page in the entire docs site. Anthropic’s prompt engineering guide is opinionated, specific, and shorter than the third-party books that paraphrase it. Reading this once is worth more than reading three “Ultimate Prompt Engineering” guides on the open web. Topics: role prompting, chain-of-thought, XML tags, few-shot examples, structured output, system vs user prompts. See our best Claude prompts for the operator-level patterns.

4. Tool use overview

URL: docs.anthropic.com/en/docs/agents-and-tools/tool-use/overview

The conceptual foundation for everything agentic. Tool use is how Claude calls functions you define — web search, database lookups, file operations, API calls. Even if you don’t write code, understanding tool use changes how you think about what Claude can actually do. See function calling glossary.

5. Vision

URL: docs.anthropic.com/en/docs/build-with-claude/vision

Claude’s image understanding capabilities. What works (charts, diagrams, screenshots, photos, handwritten notes), what doesn’t (precise OCR at small sizes, very fine spatial measurements), and how to prompt vision tasks effectively. Frequently missed by users who don’t realize Claude can see.

6. Prompt caching

URL: docs.anthropic.com/en/docs/build-with-claude/prompt-caching

The single biggest API cost-saver. Mark stable parts of your prompt as cached and Anthropic charges roughly 10% for cached tokens vs full price for fresh ones. For any application that includes a large system prompt, large reference documents, or repeated context, prompt caching cuts costs by 50-90%. Required reading for anyone shipping production code on the Claude API.

7. Batch processing

URL: docs.anthropic.com/en/docs/build-with-claude/batch-processing

The Batch API. Submit jobs that don’t need an immediate response (data enrichment, bulk summarization, long-form generation) and get them back within 24 hours at a 50% discount. Most operators ignore the Batch API and overpay for tasks that could easily run async. Worth reading even if you only sometimes need it.

8. Streaming

URL: docs.anthropic.com/en/docs/build-with-claude/streaming

Server-sent events for streaming responses. Every user-facing application should stream — perceived latency drops from “feels broken” to “feels native” the moment the first token appears. Short, focused page; required for anyone building UIs.

9. Computer use

URL: docs.anthropic.com/en/docs/agents-and-tools/computer-use

How Claude can control a virtual desktop — clicking buttons, filling forms, navigating websites — through a vision-plus-action loop. Still maturing in 2026 but already useful for QA, automation, and accessibility workflows. The closest thing to “AI doing things on your behalf in a browser” that ships from a major lab today.

10. Agent Skills overview

URL: docs.anthropic.com/en/docs/agents-and-tools/agent-skills/overview

The Skills feature. Reusable instruction packages that extend Claude’s capabilities without changing the model. See our catalog of every official Claude Skill for the practical inventory.

11. Claude Code overview

URL: docs.anthropic.com/en/docs/claude-code/overview

The official Claude Code documentation. Installation, configuration, slash commands, hooks, plugins, agents, MCP integration. Indispensable for anyone using Claude Code in real work. See our Claude Code plugins guide for the higher-leverage subset.

12. Release notes

URL: docs.anthropic.com/en/release-notes/overview

The official changelog. Models, API, Claude Code, Console. Subscribe to its RSS feed if you build anything against Claude — missing a model deprecation notice is the kind of mistake that costs production teams a weekend.

What’s not on this list and why

Several pages are valuable but specialized enough that we don’t recommend them as part of the core 12: the OpenAI-compatible API endpoint docs (useful only if migrating), the deeply technical model-card pages (useful for safety researchers), the older “Claude 2” docs (deprecated), and the dozens of language-specific SDK reference pages (useful only when you actually need them). Read the 12 first; the rest of the docs make more sense after.

How do Anthropic’s docs compare to other AI lab documentation?

Documentation siteCostBest forSource of truth
docs.anthropic.comFreeClaude API, Claude Code, Skills, prompt engineeringVendor (Anthropic)
platform.openai.com/docsFreeOpenAI API, Assistants, ChatGPT APIVendor (OpenAI)
ai.google.devFreeGemini API, Vertex AIVendor (Google)
docs.mistral.aiFreeMistral models, La PlateformeVendor (Mistral)
modelcontextprotocol.ioFreeMCP spec (cross-lab)Open spec

Anthropic’s docs are arguably the most readable of the major-lab documentation sites in 2026 — concrete examples, opinionated guidance, and prose that assumes you have a job to get done. OpenAI’s docs are dense and comprehensive. Google’s are sprawling. Anthropic’s are tight.

Key takeaways

  • Bookmark the 12 pages above. Read them in order. You’ll save dozens of hours of scattered tutorial-hunting.
  • Pages 1-5 are the everyone-needs-to-know foundation: Get Started, Models, Prompt Engineering, Tool Use, Vision.
  • Pages 6-12 are for developers and power users: Prompt Caching, Batch, Streaming, Computer Use, Agent Skills, Claude Code, Release Notes.
  • Anthropic’s docs are updated alongside every release — trust them over third-party tutorials.
  • Prompt caching and Batch API are the two most-underused cost-savers; both deserve a careful read.
  • Subscribe to the release notes RSS if you build anything production-grade.

Frequently asked questions

Are Anthropic’s docs free to access?

Yes. The entire docs.anthropic.com site is public and free. No account or API key required to read.

How often are the docs updated?

The docs update alongside every model release and feature launch — multiple times per month in practice. Major API changes get a release-note entry; model-specific updates flow into the Models page automatically.

Do I need to read all 12 pages before I can use Claude?

No. Pages 1-5 are sufficient for most consumer users. The rest are for developers and operators who want to ship production work.

What about the Anthropic cookbook?

The Anthropic Cookbook GitHub repo is a great companion — runnable example notebooks for common patterns. Read it after the docs, not instead of them. The docs explain the why; the cookbook shows working code.

Is there a single PDF version of all the docs?

No, and there shouldn’t be. The docs are updated too often for a static snapshot to stay accurate. Bookmark the pages, refresh occasionally.

What if I’m an Anthropic competitor, can I read their docs?

The docs are public. Anyone can read them. Many of the patterns Anthropic teaches are transferable to other AI APIs.

Where do I go after reading these 12?

The deeper docs (admin API, billing, prompt evaluation tools), the Anthropic Academy courses, and our operator-level prompt collection are good next steps.

You may also like

1-on-1 Coaching

Claude AI Crash Course

1-hour private video session with James. Walk through Claude Desktop, Claude Code, Cowork, Skills, Projects, file setups, and plugins. Best for owners who want a coach while rolling out workflows. No technical background required.

$75

1-hour live

Book session →

Group Format

AI Workshops for Teams

Team-format workshops for businesses rolling Claude out to staff. Best for businesses with 3+ people who all need to use the new workflows. Custom-built around your team’s actual tools and goals.

Custom

pricing

Get a quote →

Sources

Last reviewed: May 2026. All 12 URLs verified live against docs.anthropic.com.

Get Smarter About AI Every Morning

Free daily newsletter — one story, one tool, one tip. Plain English, no jargon.

Free forever. Unsubscribe anytime.

Two ways to go further

The AI Prompt Library

1,000+ ready-to-use prompts for Claude, ChatGPT, and Gemini. Stop staring at a blank box.

Get it for $39 →

2-Hour Live AI Crash Course

A private, beginner-friendly session across Claude, ChatGPT, Gemini, and the wider landscape.

Book for $125 →

Discover more from Beginners in AI

Subscribe now to keep reading and get access to the full archive.

Continue reading