12 Claude Shortcuts That Slashed My Workflow in Half. Here's the Full List.

Whether you opened Claude for the first time last week or you've been shipping with it daily since launch - there's something here you're not using.

I spent a week tracking every click, every menu navigation, every time I waited for a response I should have stopped three paragraphs ago.

51 minutes a day. Not on thinking. On clicking, scrolling, waiting, and resending messages that didn't need resending.

Then I went through every shortcut in Claude - browser and terminal. Tested each one. Timed the difference. Kept the 12 that actually changed how I work.

6 for the browser. 6 for Claude Code. All verified. All real. Here's every single one.

You can check the browser shortcuts yourself by pressing Cmd+? (Mac) or Ctrl+? (Windows) in any chat.

— BROWSER (claude ai) —

1. Cmd+K / Ctrl+K — Start New Chat

The most basic shortcut on the list.

Instead of reaching for the "Start New Chat" button in the corner, you press one combination and land in a clean input field. Context cleared. Fresh start.

Sounds trivial until you count how many times a day you start new conversations. I start 15-20. At 4 seconds saved each time, that's over a minute a day on just this one shortcut.

before: mouse to corner → click "New Chat" → wait → 4 seconds

after: Cmd+K → already typing → 0.3 seconds

The real value isn't the seconds. It's that you'll actually start fresh chats when you should instead of cramming unrelated questions into an overloaded conversation.

2. Arrow Up (↑) — Edit Last Message

This one hurts when you learn it late.

You send a prompt. Claude misunderstands. So you type "No I meant..." and send a follow-up. Now Claude re-reads the entire conversation history plus your correction. Tokens wasted. Context polluted.

Arrow up opens your last message for editing. Make sure the input field is empty first. Claude regenerates from the fixed version. The bad exchange gets replaced, not stacked.

follow-up: original + correction + full re-read = 2x tokens

edit: fixed message + clean regeneration = 1x tokens

Token math at message 20:

follow-up: ~10,500 tokens (re-reads everything + new message)

edit: ~500 tokens (just the regeneration)

21x difference. On a single message. This alone can change your monthly token spend.

3. Cmd+. / Ctrl+. — Stop Generation

Claude starts writing a massive response. Two paragraphs in you can see it's going in the wrong direction. Or it's hallucinating. Or it's writing code you didn't ask for.

Cmd+. kills it instantly. Claude keeps what it already wrote. You redirect from there.

waiting for a bad response to finish: 15-30 seconds of watching wrong output

Cmd+. at paragraph two: 0 seconds wasted

Every token Claude generates after you've seen the problem is a token you're paying for. On a long wrong response that's 500-1,000 tokens burned for nothing. I hit this 10-15 times a day.

4. Cmd+/ ; Ctrl+/ — Toggle Sidebar

The sidebar eats 280px of your screen. On a laptop that's 20% of your viewport showing a list of old conversations you aren't looking at right now.

Cmd+/ hides it. Press again - it comes back.

1440px screen with sidebar: 1160px for content

without: 1440px for content

24% more working space

Keep it closed by default. Flash it open only when switching between chats. Same logic as hiding the dock on your Mac - you don't need it visible 100% of the time.

5. Cmd+Shift+L / Ctrl+Shift+L — Toggle Dark/Light Theme

It's 11pm and the white screen is burning your retinas. Or it's morning and dark mode looks washed out in sunlight.

Cmd+Shift+L flips between dark and light mode instantly. Faster than digging through profile settings.

before: click profile → settings → appearance → toggle → 6 seconds

after: Cmd+Shift+L → done → 0.2 seconds

Minor? Sure. But you do it at least twice a day - morning and evening. And the shortcut existing means you'll actually switch when you need to instead of squinting for another hour.

6. Shift+Enter — New Line Without Sending

You're writing a structured prompt. Multiple instructions, code blocks, specific formatting. You hit Enter to make a new paragraph. The message sends. Half-finished.

Shift+Enter creates a new line. Your message stays in the input field until you actually press Enter alone.

accidental sends per day before learning this: 4-5

after: 0

Each accidental send wastes tokens on Claude's response to your incomplete thought plus the follow-up to fix it. At 4 accidents a day that's roughly 4,000 wasted tokens daily - gone because of one key.

Write your full prompt. Structure it. Then send when ready.

— CLAUDE CODE (terminal) —

7. Esc Esc — Rewind to Any Checkpoint

Single Esc stops generation. Everyone knows that.

Double Esc opens something most people have never seen - a scrollable list of every checkpoint Claude Code created during your session. Every file state. Every conversation state. Fully restorable.

Claude rewrote your auth module and made it worse? Esc Esc → restore from 3 edits ago

Claude went down a rabbit hole for 10 minutes? Esc Esc → back to before the detour

Want to try something you're only 40% sure about? Try it. Fail. Esc Esc → 5 seconds to undo everything

Four options: restore code and conversation, conversation only, code only, or summarize from checkpoint forward.

This changes how aggressively you can experiment. Zero risk. Full undo. Let Claude try wild approaches - if they fail, rewind costs nothing.

8. Ctrl+R — Reverse Search Prompt History

Claude Code keeps your full prompt history across sessions. Ctrl+R lets you fuzzy-search through all of it.

You ran a complex prompt three days ago that worked perfectly. Instead of retyping it or scrolling through terminal history, Ctrl+R → type two keywords → it's back.

retyping a 4-line prompt from memory: 45 seconds + probably wrong

Ctrl+R → "auth refactor" → exact prompt appears → 3 seconds

This gets more powerful the longer you use Claude Code. After a month your prompt history is a personal library of everything that worked. Ctrl+R is how you search it.

9. Alt+T / Option+T — Toggle Extended Thinking

Extended thinking is powerful but expensive. Adds significant tokens to every response. Most people either leave it always on or never touch it.

Option+T toggles it per message. Quick formatting question? Leave it off. Complex architecture decision? Flip it on. Back to a simple question? Off again.

thinking on a simple rename: 3,000+ tokens wasted

thinking on a complex debug: 3,000 tokens well spent

The toggle takes 0.2 seconds. The menu click path takes 4. That difference means you'll actually switch it per question instead of leaving it in whatever state it was.

10. Ctrl+G — Open Prompt in External Editor

The terminal input is one line. Writing a multi-paragraph prompt with specific instructions, code snippets, and formatting in a single terminal line is painful.

Ctrl+G opens your default text editor ($EDITOR - vim, VS Code, whatever you have set). Write your prompt in a real editor with real editing tools. Save and close. It sends.

writing a complex 10-line prompt in terminal: fighting the cursor for 2 minutes

Ctrl+G → write it properly in VS Code → save → 30 seconds

This is mandatory for anyone writing detailed CLAUDE.md-style instructions as prompts. The terminal input wasn't built for paragraphs. Your editor was.

11. Shift+Tab — Cycle Permission Modes

Claude Code asks permission for every file edit and every command. "Can I edit this file?" "Can I run npm test?" Safe but slow.

Shift+Tab cycles through three modes: normal → auto-accept → plan mode.

normal: Claude asks permission for everything (safe, slow)

auto-accept: Claude executes without asking (fast, trust required)

plan: Claude shows what it will do but doesn't execute (review first)

Switch to auto-accept when you're iterating fast on a feature you understand. Switch to plan mode when Claude is doing something you want to verify first. Switch back to normal when working on sensitive code.

One keypress to change the entire interaction speed. The mode shows below your prompt so you always know where you are.

12. /btw — Ask a Side Question Without Interrupting

Claude is mid-task. Refactoring a module. Three files deep. You suddenly need to know something - "what file was that error in?" or "what's the function signature for X?"

Before /btw: you either wait for Claude to finish (could be minutes) or Ctrl+C to cancel the whole task, ask your question, then re-prompt the original task. All context from the interrupted task - gone.

/btw lets you ask a side question while Claude keeps working. The answer comes without touching the main task's conversation history.

before: cancel task → ask question → re-prompt task → lose 5 minutes + context

after: /btw what file has the auth middleware? → answer in 3 seconds → task continues

Built by Erik Schluntz on the Claude Code team. The tweet announcing it got 2.2M views. The demand for "ask without derailing" was huge - and this is the answer.

You won't remember all 12 today. Don't need to.

Pick three that match the pain you felt today. Use them for a week. Add three more. In a month your hands will move faster than your brain can think about where to click.

Here's the full list:

BROWSER (claude. ai):

• Cmd+K — start a new chat instantly
• ↑ — edit your last message instead of sending a follow-up
• Cmd+. — stop generation mid-response
• Cmd+/ — toggle sidebar on and off
• Cmd+Shift+L — switch between dark and light theme
• Shift+Enter — new line without sending the message

CLAUDE CODE (terminal):

• Esc Esc — rewind to any checkpoint
• Ctrl+R — search your full prompt history
• Option+T — toggle extended thinking per message
• Ctrl+G — write prompts in your real editor
• Shift+Tab — cycle through permission modes
• /btw — ask a side question without interrupting a task

Mac shortcuts shown. Replace Cmd with Ctrl on Windows/Linux. Option = Alt.

I tracked a full week before and after.

before: ~51 minutes per day on clicking, navigating, waiting, resending

after: ~22 minutes on the same actions

saved: 29 minutes daily

monthly: ~10 hours

token savings from editing, stopping, and thinking toggles: ~35%

29 minutes × 250 work days = 120 hours a year.

Three full work weeks. Saved by pressing keys instead of clicking buttons.