17 hidden Claude Code commands most people never find

Ask a question without killing the run

The most expensive habit in Claude Code is stopping a working session to ask something small. You lose the run, and the answer you get costs you the 4 minutes of work you cancelled to ask it.

/btw [question]

Ask the side question and let the work continue. Run it with no question and it shows you the last one, so earlier answers stay reachable.

Official: Ask a side question about the current session without adding to the conversation.

/advisor [model|off]

It hands the hard decision to a second model before Claude commits to an approach, rather than after you have read 200 lines of the wrong one.

Official: Enable or disable the advisor tool, which consults a second model for guidance at key moments during a task.

See what you are spending, then get it back

Context is the budget. Most people never look at it, then wonder why the answers got worse an hour in. These 4 make the budget visible and give you 3 different ways to spend less of it.

/context [all]

It draws the window as a grid, so you can see what is filling it instead of guessing.

Official: Visualize current context usage as a colored grid.

/clear [name]

It's the cheapest accuracy fix there is. Old context does not sit there harmlessly, because it competes with the task in front of you.

Official: Start a new conversation with empty context.

/effort [level|auto|status|ultracode [on|off]]

Turn the reasoning down for work that does not need it. Most sessions do not need max on every turn.

Official: Set the effort level: low to xhigh, max, or auto; status prints it.

/model [model]

It's the other half of the same lever, and it sticks as your default for new sessions.

Official: Switch the AI model and save it as your default for new sessions.

Go backwards, which is the one thing chat cannot do

A chat window is a one way street. These 3 are the exits: undo a conversation, find one from last week, and give the current one a name you will recognise.

/rewind

It takes the conversation and the code back to an earlier point. Its aliases are /checkpoint and /undo, which is what most people search for first.

Official: Rewind the conversation and/or code to a previous point, or summarize from a selected message.

/resume [session]

Reopen a session by id or name, or pick from the list. The work you did on Tuesday is not gone.

Official: Resume a conversation by ID or name, or open the session picker.

/rename [name]

Name the session and the name shows on the prompt bar. This is what makes /resume usable a week later.

Official: Rename the current session and show the name on the prompt bar.

Work that keeps going when you do not

These are 3 different answers to the same question, which is what happens to a task once you close the laptop.

/loop [interval] [prompt]

It repeats a task while the session stays open, on an interval you set.

Official: Run a prompt repeatedly while the session stays open.

/schedule [description]

It moves the task off your machine entirely. It runs in the cloud on a cron, whether or not your terminal is open.

Official: Create, update, list, or run routines, which execute in the cloud.

/remote-control

It makes the running session reachable from claude.ai, so you can steer it from a phone.

Official: Make this session available for Remote Control from claude.ai.

Read your own history back

Claude Code has been keeping records of how you work. Almost nobody reads them, and these 3 do.

/insights

It writes an HTML report on your recent sessions, covering which projects, how you use it, and where things go wrong.

Official: Generate an HTML report analyzing your recent sessions on this machine: which projects you work in, how you use Claude Code, where things go wrong, and features to try.

/team-onboarding

It turns that same history into a guide for somebody joining. The onboarding doc writes itself from what you already did.

Official: Generate a team onboarding guide from your Claude Code usage history.

/fewer-permission-prompts

It reads your transcripts for the read-only calls you keep approving, then writes the allowlist so you stop approving them.

Official: Scan your transcripts for common read-only Bash and MCP tool calls, then add a prioritized allowlist to project .claude/settings.json to reduce permission prompts

The 2 that change the terminal itself

They're small, and you will notice them every day.

/statusline

It puts what you care about on the prompt bar. Context left and spend are the 2 most people choose.

Official: Configure Claude Code's status line.

/chrome

It opens the settings for Claude in Chrome. This is the one the article called an extension rather than a command.

Official: Configure Claude in Chrome settings

3 things on that list that are not commands

Naming them matters, because each one sends somebody to a prompt to type something that does nothing, and they conclude the tool is broken rather than the list.

/skill-creator

It's a skill, not a built-in. You install it before you can run it, and it does not appear in the command reference.

/caveman

It's also a skill, and an open-source one. Claude Code doesn't ship with it, but it's worth installing.

ULTRATHINK

It's a word you type in a prompt, not a command. The built-in that sets reasoning depth is /effort.

The other 101

17 is where to start, not where it ends. The full reference is on all Claude Code commands, every one of them read from the same source as this page.

More AI coding tools I've reviewed

These are my reviews of Claude Code and the agents I compare it with:

Claude Code

Paid

AI coding agent that also writes, researches, and runs full business workflows from plain English.

Code assistantsAI agentsLLMProductivity

Codex

Paid

OpenAI's agentic coding assistant that writes, edits, and ships code from natural-language instructions.

Code assistantsAI agentsProductivity

Cursor

Freemium

Code editor with AI pair-programming and intelligent autocomplete.

Code assistantsProductivity

Final thoughts on hidden Claude Code commands

A wrong answer reads exactly like a right one, which is why every command here was checked against the reference first.

Keep reading:

• Claude Code commands – all 118, grouped by when you reach for one
• Claude Code permission modes – the 6 modes, and the keypress that switches them
• What is Claude Code? – what it does, and whether it is for you
• How to install Claude Code – every route, and which ones update themselves

Then try /btw the next time you want to stop a run to ask something small. And when you'd rather write a better prompt than stop a run, my Claude Code prompt generator builds one for you.