Claude 101
Ten questions that account for most of the confusion around Claude, answered in plain language — then the route through the rest of the reference for whoever you happen to be.
This is the 101. It is not where the reference stops: the same site covers model retirement schedules, enterprise controls, rate-limit tiers and system cards.
Looking for Anthropic's own course of the same name? That isClaude 101 on Academy — 13 lessons, two and a half hours, and a different job from this page.
The short version
Ten things worth knowing before anything else
- 01
What is Claude?
An AI assistant from Anthropic, used through a web app, a desktop app and mobile apps. Everything else with "Claude" in the name is a separate product built around it.
Read the full entry → - 02
Why are there so many products?
Because they do different jobs. Claude answers; Claude Code changes a codebase; Cowork produces a deliverable from your files. Same models underneath, very different postures.
Read the full entry → - 03
Which model should I use?
Start with Claude Opus 5 — that is Anthropic’s own default recommendation. Try Sonnet 5 for production work; it is 60% cheaper and usually holds up.
Read the full entry → - 04
What do the model names mean?
Fable, Opus, Sonnet and Haiku are capability tiers, most to least. The number is the generation. Fable 5 is the most capable widely released model; Haiku 4.5 is the fastest and cheapest.
Read the full entry → - 05
Which plan do I need?
Free covers the assistant. Pro is the threshold that unlocks Claude Code, Cowork, Design and Science all at once — that, not the extra usage, is usually the real decision.
Read the full entry → - 06
Is a subscription the same as API access?
No, and this trips up almost everyone. They are separate products with separate billing, separate limits and separate accounts. Having one grants nothing on the other.
Read the full entry → - 07
What is MCP?
An open protocol that lets Claude reach systems outside itself — your database, your ticket tracker, your files — through one standard interface instead of a bespoke integration each time.
Read the full entry → - 08
What are Skills?
Folders of instructions Claude loads only when a task calls for them. Skills are knowledge; MCP is access. Most confusion in this ecosystem is one being used to solve the other’s problem.
Read the full entry → - 09
Will my setup keep working?
Not automatically. Every model ID is a pinned snapshot, models get retired on a published schedule, and requests to a retired model fail outright.
Read the full entry → - 10
Where do I go when something changed?
Here. Every page records when it was last verified and against what, and material changes are logged with dates.
Read the full entry →
Then take a route
The 101 ends differently depending on who you are
Just curious
Read what Claude actually is, then look at what separates it from Claude Code and Cowork.
A developer
Claude Code is the product you want. Learn what it reads at startup and what it is allowed to do.
Building on the API
The Messages API is the whole surface. The rest is parameters, caching and knowing what got deprecated.
Rolling it out at work
The plan decides the controls. Team gives you SSO and billing; Enterprise gives you audit logs and SCIM.
The six entries most people need first
Claude
GAAnthropic's assistant, available as a web app, desktop app and mobile app, and the surface most people mean when they say "Claude".
Products
Claude Code
GAAnthropic's agentic coding product — it reads a codebase, edits files, runs commands and integrates with development tools, across terminal, IDE, desktop, web and CI.
Products
Claude Cowork
GAAnthropic's knowledge-work agent — you give it a goal, and it works across your local files and connected apps to return a finished deliverable.
Products
Model Context Protocol
GAAn open standard that lets an AI application connect to external tools and data through one client/server interface instead of a bespoke integration per tool.
Agent Stack
Agent Skills
GAFolders of instructions, scripts and resources that Claude loads on demand when a task calls for them — packaged, portable expertise rather than a longer prompt.
Agent Stack
Pro
GAThe entry paid plan, and the point at which Claude Code, Cowork, Design and Science become available.
Plans & Limits
One thing this page will not do
Teach you to use Claude. Anthropic's ownAcademy does that properly and for free —AI Fluency is four hours and worth all of them, and The 4 Properties of AI is seven minutes.
What this reference does instead is answer the questions the documentation is not organised to answer: how these products differ, what your plan actually includes, and whether what you read six months ago is still true. Every claim here carries a date and a source.