13 Claude Code Commands Every Beginner Should Know (With Real Examples)

Make Claude remember facts/preferences in CLAUDE.md.

When to use it: To persist project-specific rules across sessions.

Switch AI model (e.g., Sonnet to Haiku).

When to use it: When you need faster responses or deeper reasoning.

Undo and roll back conversation or code changes.

When to use it: If the AI takes a wrong turn and you need a reset.

See a step-by-step plan before Claude acts.

When to use it: For complex refactors where you need to verify intent.

Set a finish condition and let Claude work autonomously.

When to use it: When you want Claude to complete a task without hand-holding.

Create specialized subagents for specific tasks.

When to use it: To delegate parts of a project to isolated expert instances.

Run a task in the background while you keep working.

When to use it: For long-running tests or data processing.

See how full the context window is.

When to use it: To prevent performance drops due to context bloat.

Summarize conversation to free up space.

When to use it: During long sessions to keep Claude focused and cheap.

Check session cost and plan limits.

When to use it: To stay within budget and monitor token consumption.

Create a safe copy of the conversation to try risky changes.

When to use it: Before experimenting with destructive code changes.

Scan pending changes for security issues before shipping.

When to use it: Final check before committing sensitive code.

Run a prompt repeatedly on an interval.

When to use it: For monitoring logs or waiting for a specific build state.