The Five Claudes You're Not Using

Open your browser history. If every Claude visit for the past month points at claude.ai/chat, this post is about you.

The chat window is the front door. It's the cheapest thing Anthropic can ship — a text box, a send button, one context. The other five surfaces are where most of the work your team bills for actually lives. This is a diagnostic, then a tour, ordered by leverage rather than by Anthropic's marketing diagram.

The diagnostic

One question per surface. Count your yeses.

1. Do you retype the same "act as our brand voice, use these rules, output in this format…" preamble more than twice a week?
2. Do you paste a single file into chat and wish it could see the other forty?
3. Do you re-explain a long-running client, codebase, or research topic because a chat went cold?
4. Do you download a file, rename it, drop it in a folder, and repeat that twenty times?
5. Do you ask Claude to "describe" a deck, then go rebuild it by hand in Slides?

Every "yes" is a surface you're paying for in human hours. Six yeses means you're using one-sixth of what's already included in your subscription.

The five surfaces, ordered by leverage

I'm not listing these the way the official diagram does. The ordering below is what actually moves the needle once you commit to adopting one per week.

1. Claude Skills — the highest-leverage surface, and the one most people don't know exists

What it is: a custom instruction set — scoped rules, format, examples — that Claude loads automatically when a matching task shows up. A function you define once and call by intent instead of by name.

The shift: from "paste the preamble every time" to "the preamble is permanent."

Concrete example: a skill called write-release-notes that pulls from your changelog format, enforces the three-section structure Marketing approved in Q1, and always writes in past tense. You never type "use our release notes format" again. It just happens.

When not to use it: for one-off tasks. A skill you invoke once a quarter is worse than a saved prompt — it's a skill you'll forget exists and rewrite badly in six months.

2. Claude Code — the one that replaces half your IDE

What it is: a terminal-native coding agent with file-system memory, the ability to run tests, refactor across files, and build multi-file features end-to-end.

The shift: from "paste a file, ask for help, paste the edit back" to "describe the change, watch it land, review the diff."

Concrete example: "Rename UserProfile to Account across the codebase, update the prisma schema, write the migration, update tests." One command. One review.

When not to use it: greenfield exploration where you haven't decided on architecture yet. Code is a senior engineer with a task list, not a collaborator for "what should we build?" For that, the chat surface still wins.

3. Claude Projects — persistent context for the work that outlives a single chat

What it is: a workspace that remembers your reference files, custom instructions, and conventions across every conversation inside it.

The shift: from "re-explain the client every Monday" to "the client is the room."

Concrete example: one Project per client. Drop their brand guide, tone-of-voice doc, last three decks, and working glossary into it. Every conversation inside the project inherits that context without you reciting it.

When not to use it: fast, exploratory research with no long-lived reference material. Spinning up a Project for a one-hour task is premature organization.

4. Claude Cowork — file-system automation on your actual desktop

What it is: a desktop app that can read, move, rename, extract, and transform files across your Mac or Windows machine. The agent doesn't just write text — it touches your file system.

The shift: from "export, rename, drop, repeat" to "organize this folder by project and year, extract the invoice totals into a CSV."

Concrete example: thirty PDF invoices in a Downloads folder become a tidy Clients/2026/ tree with a summary CSV. The kind of task you've been doing manually every month.

When not to use it: anything involving sensitive files you don't want an agent touching without review. Desktop access is power; treat it like sudo.

5. Claude Design — the visual surface nobody saw coming

What it is: a workspace for generating polished decks, one-pagers, wireframes, and mockups directly from prompts and reference material — not descriptions of slides, actual slides.

The shift: from "Claude drafts the copy, you rebuild the deck" to "Claude ships the deck, you edit it."

Concrete example: "Turn this product-spec doc into a six-slide exec update with our brand colors, three charts, and a one-slide risks section." Output: an editable deck.

When not to use it: anything where your design system is complex and central to brand equity. Design is a draft surface — it accelerates the first 80%, not the last 20% a senior designer does.

(0) Claude Chat — the one you already know

You already know this one. The mistake is staying here for everything. Chat is for thinking out loud and one-off synthesis. The moment you're pasting the same preamble twice, you've outgrown it.

The decision rule: which surface do you open next Monday?

Don't adopt all five at once. Pick one based on what you did most of last week:

Use it for a week. If it stuck, keep it and move to the next one. If it didn't — and sometimes it won't — close the tab and try another. The point is not to use all six; it's to stop using the one that doesn't fit the job.

The honest caveat

Anthropic adds and renames these surfaces faster than any blog post can keep up. Names will shift, capabilities will overlap, and at least one of the six will probably be folded into another within a year. What won't change: the chat window is the smallest surface, and the teams getting the most out of Claude are the ones who stopped treating it as the only one.

If your team has a "Claude power user," ask them which surface they opened last. If the answer is still the chat window, there's a week of leverage waiting.