Since there are already plenty of tutorials on Claude Code, I decided to create a quick reference guide with all the info I’ve used so far.
Starting a Session
| Command | Description |
|---|---|
claude | Start a new interactive session |
claude -c | Resume the previous session |
claude --resume | Pick a past session to resume from a list |
claude "your task" | One-shot query, no interactive session |
claude -p "query" | Print mode (non-interactive, good for piping) |
claude update | Update Claude Code to latest version |
Permission Modes (at startup)
| Command | Description |
|---|---|
claude --permission-mode plan "task" | Read-only — Claude plans but touches nothing |
claude --permission-mode acceptEdits "task" | Auto-accepts file edits, still asks for bash commands |
claude --dangerously-skip-permissions "task" | Full auto — no confirmations at all (YOLO mode) |
Models
| Command | Description |
|---|---|
claude --model claude-sonnet-4-6 | Use Sonnet (default, fast, cheaper) |
claude --model claude-opus-4-6 | Use Opus (slower, more powerful, 5x cost) |
/model | List available models inside a session |
Plugins & MCP
| Command | Description |
|---|---|
claude plugin install code-simplifier | Install official plugin |
claude mcp add mariadb -- <command> | Add a MCP server |
/mcp list | List active MCP servers inside a session |
claude mcp add my_server \
-e MYSQL_HOST="192.168.1.3" \
-e MYSQL_PORT="3306" \
-e MYSQL_USER="root" \
-e MYSQL_PASS='root' \
-e MYSQL_DB="my_database" \
-- npx @benborla29/mcp-server-mysql
claude mcp add mongodb /
-e MDB_MCP_CONNECTION_STRING='mongodb://admin:admin@192.168.1.3:27018/my_database' /
-- npx -y mongodb-mcp-server
Inside a Session (slash commands)
| Command | Description |
|---|---|
/context | Show the current usage of Context Window |
/help | List all available slash commands |
/memory | View and edit Claude’s memory files |
/config | Open settings menu |
/compact | Compress conversation history to save tokens |
/clear | Clear the screen |
/init | Bootstrap a CLAUDE.md from your codebase |
/export full_history.md | Export full conversation to a Markdown file |
/plugin install <name> | Install a plugin from inside a session |
Inside a Session (keyboard shortcuts)
| Shortcut | Description |
|---|---|
Shift+Tab | Cycle permission modes: default → acceptEdits → plan |
Ctrl+G | Open Claude’s plan in your editor to review/edit |
Option+Enter | New line without sending (multi-line input) |
Ctrl+C | Stop current generation |
Ctrl+D | Exit Claude Code |
File Input / Output
| Syntax | Description |
|---|---|
@specs/file.md | Load a file as input context |
"write output to plan.md" | Ask Claude to write output to a file |
My Workflow Per Feature
# 1. Plan
claude --permission-mode plan "@specs/02-feature.md write a plan.md"
# 2. Review plan.md in editor (Ctrl+G or open manually)
# 3. Implement
claude --dangerously-skip-permissions -c "implement plan.md exactly as written"
git add . && git commit -m "feat: feature name"