Skip to main content

Hello Claude: Your First Conversation

You've installed Claude Code your General Agent. Now it's time to actually use it.

This lesson isn't about learning complex features. It's about having your first real conversation with Claude and realizing it's not intimidating—it's actually natural.

The Setup: One Command​

Open your terminal and type:

claude

That's it. You're now in Claude Code.

You'll see a prompt. It looks something like:

▗ ▗   ▖ ▖  Claude Code v2.1.9                                                                                                                                              
           Haiku 4.5 · Claude Max                                                                                                                                          
  ▘▘ ▝▝    ~/Documents/code/panaversity-official/tutorsgpt/mem                                                                                                   
> _

This is where you talk to Claude. In natural language. No special syntax. No commands you need to memorize.

Your First Questions: Orient Yourself​

Let's start with the basics. Ask Claude where you are:

> Where am I right now??

Claude will tell you. Your current directory. Simple context.

Now ask what Claude can do:

> What can you do? What are my main options?

Claude will explain it naturally. Not a manual. Not technical jargon. Just what's possible.

What you're learning: Claude responds to natural language. You're not "commanding"—you're asking questions and conversing.

Real-World Example: Search for News and Save It​

Here's something useful. Let's find recent AI news and save it to a file.

Ask Claude:

> Search for the latest news about AI agents in 2026 and create a file called 'ai-news-2026.txt' with what you find

Watch what happens:

  1. Claude reads the web for recent news
  2. Claude proposes a file with formatted content
  3. You'll see a prompt asking: > Create file ai-news-2026.txt?
  4. You press Enter to approve

Done. You now have a file with actual, current information.

What just happened:

  • You described what you wanted in plain English
  • Claude figured out the steps (search, format, create file)
  • Claude showed you what it was about to do
  • You approved with one keypress

No clicking. No menu hunting. Just conversation.

The Approval Pattern: You're Always in Control​

Notice that Claude didn't just do things. It asked first.

This is the permission model. Every time Claude wants to:

  • Read a file
  • Write or create a file
  • Run a command

It shows you exactly what it's about to do and waits for your approval.

Press [Enter] to approve. Press [Esc] to reject.

If you reject, you can ask Claude to try differently. That's steering.

Why this matters: You're never surprised. Claude can't accidentally delete your project or run something dangerous without your knowledge.

Next: Extending Claude (Preview)​

Once you master basic conversations, you'll start asking more ambitious questions:

  • "How do I make Claude repeat this task every morning?"
  • "How do I connect Claude to my Jira board so it can check for new tickets?"
  • "How do I enforce my team's naming conventions automatically?"
  • "How do I give Claude access to my company's database?"

This is where Claude's extension tools come in. You don't need to understand them yet—just know they exist.

The Five Extension Tools at a Glance​

If you need...Then use...Appears in...
Persistent context for every Claude sessionCLAUDE.mdLessons 5-6
Repeated procedures done your waySkillLessons 7-8
Delegation & focus for complex workflowsSubagentLessons 11-12
External systems & data (APIs, databases, websites)MCPLessons 9-10
Automated quality gates & safety rulesHookLesson 13

Here's the pattern: You learn each tool individually (Lessons 5-13), then in Lesson 16, you'll see how an expert combines them in a production workflow—like a manager building a team, not a developer writing functions.

Why This Matters Now​

As you continue through the chapter, you'll recognize patterns:

  • "I need Claude to always format code like this" → You'll think: That's a CLAUDE.md use case
  • "I want Claude to research and write a report without interrupting me" → That's a Subagent
  • "I need Claude to check GitHub for vulnerabilities before I commit" → That's a Hook

You're building a mental model. The tools are the same; the architecture is what separates amateur automation from professional systems.

Try With AI​

Explore naturally:

"Ask Claude: 'What's in this directory?' Then ask: 'Tell me about my project structure.' Watch how Claude learns about your environment through conversation."

Create something real:

"Ask Claude to search for recent breakthroughs in your field of interest and save them to a file. Use natural language—describe what you want, not how to do it."

Test the approval workflow:

"Ask Claude to create a file, and when it asks for approval, press Esc to reject. Then ask it to do something different. Notice how Claude adapts without frustration."

Next Up: Now that you're comfortable conversing with Claude, let's teach Claude about YOUR preferences.