The menu bar of a program you cannot click. In Word you open the File menu with the mouse. In Claude Code there is no mouse, so the menu opens when you type /.

A slash command is an instruction to the harness, not to the model. When you type /clear, no AI is involved. The program itself does something: wipes the conversation, shows a picker, changes a setting. Everything you type without a slash goes to the model as a prompt.

Type / on an empty line and the full menu appears. Keep typing to filter it: /mo leaves /model. Press Enter to run the highlighted one, Esc to close the menu.

The ten to know first

TypeDoes
/helpLists every command and shortcut.
/clearStarts a new session in the same folder. The old one is saved.
/resumeShows your saved sessions in this folder so you can pick one up again.
/modelPicks the model tier. See Claude model family.
/effortSets how long the model thinks before it answers. See Effort.
/contextShows how full the context window is and what is in it.
/usageShows how much of your five-hour and weekly allowance is left. See Usage limits. /cost is the same command.
/memoryLists your instructions files and opens the one you pick.
/statuslineSets up a status line under the prompt in plain English. See Set up a status line.
/exitLeaves Claude Code.

The full list, with the ones you will meet later, is on the Claude Code cheat sheet.

Example

You have been working on a report for an hour and want to switch to sorting invoices. You type /clear. The whiteboard is wiped and the invoices get a clean start. Tomorrow you type /resume, pick the report session from the list, and carry on where you left off.

Why it matters for you

Slash commands are how you steer the program around the model: which brain, how much it thinks, how full the whiteboard is, when to start fresh. Once /clear and /model are in your fingers, most of the “the AI is being strange” moments have a one-word fix. And when you forget one, / shows you the whole menu.