Claude Code keeps a per-directory history of the prompts you've typed, searchable the same way your shell history is.

What it does

Every prompt you submit gets stored, scoped to the project directory. Up and down arrows cycle through entries. Ctrl+R opens reverse search - type a fragment and Claude jumps to the most recent match. History survives restarts, so repeated prompts like "run the tests" are one keystroke away.

When to use it

Re-running common prompts without retyping them.
Grabbing a complex prompt you wrote yesterday and tweaking it.
Building a mental library of what works in a given repo.
Quickly returning to a prompt after a failed attempt.

Gotchas

History is per-directory. Switching projects starts fresh.
There's no built-in secret redaction - if you paste an API key into a prompt, it sits in history. Clear it manually.
Very long prompts still take a single slot. Reverse search only matches the first line well.

Official docs: https://code.claude.com/docs/en/interactive-mode.md#command-history