Claude Code slash commands

Shows every slash command with a one-line description. Run this first in any new session. Also shows any custom project commands defined in .claude/commands/.

Wipes the current session's conversation history and starts fresh — same project context, new conversation. Use when you want a clean slate without closing and reopening. Your rate limit usage is unchanged.

Condenses the conversation into a compact summary, freeing context window space while preserving your current task state. Optionally pass instructions to guide what the summary emphasizes:

/compact focus on the auth module changes

Run before switching to a new subtask, or when responses start feeling imprecise from a long context. Does not affect your rate limit meters.

→ Full /compact guide

Exits Claude Code. Same as Ctrl+C or closing the terminal window.

Prints your current usage against both rate limit windows:

Session (5h):  34% used · resets in 2h 44m
Weekly (7d):   61% used · resets in 2d 7h

Same numbers as ~/.claude/headroom-usage.json — but you have to remember to run it. If you want both meters visible at all times without a command, Headroom shows them in the menu bar, color-coded before a limit stops you.

→ Rate limits explained · 5-hour window · 7-day window

Breaks down input tokens, output tokens, cache reads, and the estimated dollar cost for the current conversation. Useful for spotting expensive patterns — a large file being re-read every turn, or a long context that should have been compacted.

/cost
# Input tokens:  48,234
# Output tokens:  3,891
# Cache reads:   42,100
# Estimated cost: $0.42

→ Cost tracking guide

Without arguments, shows the model in use. With a model ID, switches immediately:

/model
# Current model: claude-sonnet-4-6

/model claude-opus-4-8
# Switched to claude-opus-4-8

The switch persists for the rest of the session. To set a permanent default, use "model" in ~/.claude/settings.json.

→ Model selection guide

Opens an interactive configuration menu for common settings: model, theme, verbosity, and more. Changes are written to ~/.claude/settings.json.

Shows what Claude Code has stored in memory — facts about the project, your preferences, and decisions made in previous sessions. You can review and delete specific memories.

/memory
# Lists all stored memory items with IDs

/memory delete abc123
# Removes that memory item

Runs a code review on uncommitted changes or a specified set of files. Looks for bugs, security issues, and style problems. Can be used before committing to catch issues early.

/review
# Reviews git diff HEAD

/review src/auth.ts
# Reviews a specific file

Shows all Model Context Protocol servers currently connected, their status, and which tools they expose. If a server shows as disconnected, restart it with /mcp restart <name>.

/mcp
# • filesystem  ✓ connected  (tools: read_file, write_file, list_directory)
# • github      ✓ connected  (tools: get_issue, list_prs, get_file)
# • postgres    ✗ disconnected

→ MCP setup guide

Any .md file in .claude/commands/ (project-level) or ~/.claude/commands/ (user-level) becomes a slash command. The filename is the command name; the file content is the prompt template.

# .claude/commands/deploy.md
Deploy the app to staging: run `npm run build`, then `railway up --ci`,
then verify with `curl https://staging.example.com/health`.

Run it with /deploy. Use $ARGUMENTS in the template to pass dynamic content: /deploy production.