Skip to main content

Core Commands, Custom Commands & Workflows

From Setup to Daily Use: Mastering Claude Code Commands​

In Lesson 2, you installed Claude Code and saw it work for the first time. Now comes the crucial step: learning the commands you'll use every day to work efficiently with your AI pair programmer.

This isn't just about memorizing syntax. It's about understanding when to use each command, why it matters, and how to build workflows that make development faster and more enjoyable.

Command Reference Table​

These are the commands you'll use most often. Don't try to memorize them all at once—use this table as a reference, then practice in the hands-on section.

CommandPurposeWhen to UseExample
claudeStart a sessionBegin a new task or ask a questionclaude
#Save a MemorySave key decisions in CLAUDE.md# We will use python with type hinting
@filenameReference a specific fileGive Claude context about a file@app.py what does this function do?
/initSet up project memoryTell Claude about your project (creates CLAUDE.md)/init
/clearStart freshEnd current conversation, start new task/clear
/compactSummarize conversationWhen approaching token limit/compact
ESCStop generationClaude is generating too much, stop itPress ESC once
ESC ESCEmergency stopClaude won't stop, force quitPress ESC twice quickly
/mcpManage external toolsList or add MCP servers/mcp list
/usageCheck usageShow plan usage daily/weekly limits/usage
/permissionsControl accessSet what Claude can do/permissions

Core Commands (Explained)​

1. claude - Start a Conversation​

Purpose: The main command to interact with Claude Code

Syntax:

claude                           # Start interactive conversation
claude "your prompt here"        # One-off request

When to use:

  • Starting a new task
  • Asking a question
  • Getting help with an error

Example:

claude "Review the authentication logic in auth.py and suggest improvements"

What happens: Claude reads the file, analyzes the logic, and provides specific suggestions

2. # - Create Checkpoints​

Purpose: Mark progress points in long conversations

Syntax:

# Your checkpoint message here

When to use:

  • After completing a subtask
  • Before starting a new feature
  • To organize multi-step work

Example:

# Fixed the login bug
# Now working on password reset

Why it matters: Checkpoints help Claude (and you) track what's done vs. what's next. In long conversations, they prevent Claude from losing context.

3. @filename - Reference Files​

Purpose: Tell Claude to look at a specific file

Syntax:

@filename.ext your question

When to use:

  • Asking about a specific file
  • Comparing multiple files
  • Getting file-specific help

Example:

@models.py @views.py how do these two files interact?

What happens: Claude reads both files and explains their relationship

4. /init - Set Up Project Memory​

Purpose: Initialize Claude Code for your project (creates CLAUDE.md)

Syntax:

/init

When to use:

  • First time using Claude Code in a project
  • When you want Claude to remember project context

What it does:

  1. Asks about your project (name, language, purpose)
  2. Creates CLAUDE.md with your answers
  3. Claude reads this file in future sessions

Time saved: 2-5 minutes per session (no more repeating project context)

5. /clear - Start Fresh​

Purpose: End current conversation, clear context

Syntax:

/clear

When to use:

  • Switching to a completely different task
  • Current conversation is getting confusing
  • Want to start over without previous context

Example scenario:

# Working on feature A
claude "Help me with authentication"
# ... conversation ...

/clear

# Now working on feature B
claude "Help me with database migration"

What gets cleared: Conversation history, file references What stays: Your project (files), CLAUDE.md

6. /compact - Summarize Conversation​

Purpose: Condense long conversation to save tokens

Syntax:

/compact

When to use:

  • Long conversation approaching token limit
  • Want to keep context but reduce size
  • Conversation has lots of back-and-forth

What it does:

  • Claude summarizes the conversation
  • Keeps key decisions and progress
  • Removes redundant exchanges

vs. /clear: /compact keeps context (summarized), /clear removes all context

7. ESC and ESC ESC - Stop Generation​

Purpose: Interrupt Claude when it's generating

When to use:

  • Claude is generating too much detail
  • Going in wrong direction
  • Just want it to stop

ESC (once): Polite stop - Claude finishes current thought and stops ESC ESC (twice): Force quit - immediate stop

Example:

claude "Explain dependency injection in detail"
# Claude starts generating paragraphs...
# Press ESC - Claude stops gracefully

8. /mcp - Manage External Tools​

Purpose: List, add, or configure MCP servers (external tools)

Syntax:

/mcp list                                    # Show connected tools
/mcp add <name> <command>                   # Add a tool

When to use:

  • You want Claude to access external data (web, databases, APIs)
  • Need to check what tools are connected
  • Setting up MCP servers (covered in Lesson 6)

Example:

/mcp list

Output: Shows all connected MCP servers (e.g., web search, docs, GitHub)

9. /usage - Check API Usage (Budget Awareness)​

Purpose: Show plan usage limits

Example Output: Settings: Status Config Usage (tab to cycle)

Current session ███ 6% used Resets 2pm (Asia/Karachi)

Current week (all models) ███████████ 22% used Resets Nov 12, 10pm (Asia/Karachi)

Current week (Opus) 0% used

Esc to exit


**Best practice**: Check `/usage` often to prevent unexpected session limit reached problems.

---

### 10. `/permissions` - Control Access (Security)

**Purpose**: Set what Claude Code can access and do

**Syntax**:
```bash
/permissions                      # Show current permissions
/permissions set <permission>    # Change a permission

When to use:

  • First time setting up (define boundaries)
  • Working with sensitive files
  • Want to restrict Claude's actions

Example:

/permissions

Output:

Current permissions:
- Read files: ✓ Allowed
- Write files: ✓ Allowed (approval required)
- Execute commands: ✓ Allowed (approval required)
- Access network: ✗ Denied

Custom Slash Commands (Team Workflow Automation)​

What Are Custom Slash Commands?​

Custom slash commands are reusable prompt templates you create for tasks you do repeatedly.

Think of them as: Shortcuts for common workflows

Stored in: .claude/commands/ directory in your project

How to Create a Custom Command​

Step 1: Create the commands directory

mkdir -p .claude/commands

Step 2: Create a markdown file for your command

Example: .claude/commands/markdown-review.md

# Markdown Review Command

Review the Markdown in $ARGUMENTS for:
1. Potential bugs or edge cases
2. Style and readability
3. Security vulnerabilities

Provide:
- Summary of findings (critical/major/minor)
- Specific line numbers for issues
- Suggested fixes with examples

Format: Use a clear severity rating for each issue.

Step 3: Use your custom command

claude /markdown-review >> We are learning how to Create a custom command

What happens:

  • $ARGUMENTS gets replaced with >> We are learning how to Create a custom command
  • Claude executes the full prompt template
  • You get consistent, structured reviews

Why Custom Commands Matter​

Without custom commands:

claude "Review auth.md for bugs, style issues, performance problems, and security vulnerabilities."

**With custom commands**:
```bash
claude /markdown-review `>> We are learning how to Create a custom command`

(One short command)

Benefits:

  • ⏱️ Time saved: 30 seconds per use
  • 🎯 Consistency: Same quality checks every time
  • 👥 Team alignment: Share commands via Git (everyone uses same prompts)
  • 📚 Best practices: Encode expertise into commands

Try With AI​

Use Claude Code CLI for this activity (since you're learning Claude Code commands).

Prompt 1: Command Selection​

I'm working on [describe your task]. Which Claude Code commands should I use, and in what order? Give me a step-by-step workflow with specific commands. Include when to use checkpoints, when to check /cost, and when to /clear or /compact.

Expected outcome: Step-by-step command workflow tailored to your task

Prompt 2: Custom Command Creation​

I do this task repeatedly: [describe repetitive task]. Help me create a custom slash command for it. Write the markdown file content I should save in .claude/commands/[name].md. Make it use $ARGUMENTS so I can pass different files/inputs.

Expected outcome: Ready-to-use custom command file content

Prompt 3: Workflow Debugging​

I tried the [Explore→Plan→Code / TDD] workflow but got stuck at [step]. What went wrong? Walk me through that step again with specific commands and what I should see at each point.

Expected outcome: Troubleshooting help with exact commands to try

Prompt 4: Usage Optimization​

I just ran /usage and my usage is higher than expected. Review my recent workflow: [describe what you did]. Which commands or approaches would reduce token usage while keeping the same quality? Give me 3-5 specific optimization tips.

Expected outcome: Practical tips to optimize usage