ChatGPT API Tutorial: Build Your First AI-Powered App in 1 Hour

ChatGPT API Tutorial: Build Your First AI-Powered App in 1 Hour

Using ChatGPT via the web interface has a ceiling. The API removes it.

With the API, you build AI into your own applications — custom interfaces, automated workflows, AI features in products you're building, or scripted processes that run without a human typing prompts.

I'll walk you through everything: account setup, your first API call, understanding the cost model, and a complete working application by the end of the tutorial.

Prerequisites: Basic programming knowledge in Python or JavaScript. No AI experience required.

Step 1: Account Setup and API Key

1. Go to platform.openai.com and create an account (separate from your ChatGPT account, though you can use the same credentials)
2. Navigate to Settings → Billing and add a payment method
3. Optionally set a usage limit in Billing — I recommend $10–$20 for learning to prevent unexpected charges
4. Go to API Keys → Create new secret key
5. Copy the key immediately — you won't be able to see it again

Security: Never commit API keys to version control. Store them in environment variables.

Step 2: Install the OpenAI Library

Python:

pip install openai

Node.js:

npm install openai

Step 3: Your First API Call

Python:

from openai import OpenAI

client = OpenAI(api_key="your-api-key-here")  # Or use os.environ["OPENAI_API_KEY"]

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What are three tips for learning to code faster?"}
    ]
)

print(response.choices[0].message.content)

Node.js:

import OpenAI from "openai";

const client = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });

const response = await client.chat.completions.create({
  model: "gpt-4o-mini",
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "What are three tips for learning to code faster?" }
  ]
});

console.log(response.choices[0].message.content);

Run this and you'll get an AI-generated response. That's the core of the API.

Understanding the Response Structure

The API returns an object with this structure:

{
  "id": "chatcmpl-...",
  "choices": [{
    "message": {
      "role": "assistant",
      "content": "Here are three tips..."
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 28,
    "completion_tokens": 145,
    "total_tokens": 173
  }
}

Key parts:

• choices[0].message.content — the actual response text
• usage.prompt_tokens — tokens in your input
• usage.completion_tokens — tokens in the response
• usage.total_tokens — total (what you're billed for)

Understanding Tokens and Pricing

What's a token? Roughly 4 characters of English text. "Hello, world!" is about 4 tokens. A 1,000-word article is roughly 1,300 tokens.

Model pricing (approximate, as of 2026):

Cost example: A chatbot processing 1,000 conversations per day, each averaging 500 input + 300 output tokens:

• gpt-4o-mini: ~$0.21/day
• gpt-4o: ~$7.00/day

For most applications, start with gpt-4o-mini and upgrade only if quality isn't sufficient.

The System Prompt: Your Most Powerful Tool

The system prompt defines how the model behaves. It's the difference between a generic AI and an AI that acts like your specific assistant.

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {
            "role": "system",
            "content": """You are a customer service agent for TechStore, an electronics retailer. 
            Only answer questions related to our products and policies. 
            If asked about something unrelated, politely redirect to relevant topics.
            Always be helpful and professional.
            Our return policy: 30-day returns on all items with original packaging."""
        },
        {"role": "user", "content": "Can I return headphones I bought 3 weeks ago?"}
    ]
)

This single system prompt makes the model behave as a constrained, company-specific assistant.

Building a Complete App: AI-Powered Content Summarizer

Let's build something useful: a script that reads an article URL, fetches the content, and generates a summary.

Python (complete working script):

import os
from openai import OpenAI
import requests
from bs4 import BeautifulSoup

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

def fetch_article(url):
    """Fetch article text from URL"""
    response = requests.get(url)
    soup = BeautifulSoup(response.text, 'html.parser')
    paragraphs = soup.find_all('p')
    text = ' '.join([p.get_text() for p in paragraphs])
    return text[:4000]  # Limit to first 4000 chars to manage tokens

def summarize_article(text, summary_type="brief"):
    """Generate article summary using ChatGPT API"""
    
    if summary_type == "brief":
        instruction = "Summarize this article in 3 bullet points. Be concise."
    elif summary_type == "detailed":
        instruction = "Provide a detailed summary: main topic, key arguments, and conclusion."
    elif summary_type == "tldr":
        instruction = "Write a one-sentence TL;DR for this article."
    
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[
            {
                "role": "system",
                "content": "You are a skilled editor who creates accurate, concise article summaries."
            },
            {
                "role": "user",
                "content": f"{instruction}\n\nArticle text:\n{text}"
            }
        ],
        max_tokens=500
    )
    
    return response.choices[0].message.content

# Main
if __name__ == "__main__":
    url = input("Enter article URL: ")
    print("Fetching article...")
    
    article_text = fetch_article(url)
    
    print("\n=== BRIEF SUMMARY ===")
    print(summarize_article(article_text, "brief"))
    
    print("\n=== TL;DR ===")
    print(summarize_article(article_text, "tldr"))
    
    print(f"\nTokens used: approximately {len(article_text.split()) * 1.3:.0f}")

Dependencies: pip install openai requests beautifulsoup4

This is a complete, working application. Run it, paste in a news article URL, and get a summary.

Multi-Turn Conversations

The API is stateless — each call is independent. To build a conversation, you maintain the message history yourself:

conversation_history = [
    {"role": "system", "content": "You are a helpful cooking assistant."}
]

while True:
    user_input = input("You: ")
    if user_input.lower() == "quit":
        break
    
    # Add user message to history
    conversation_history.append({"role": "user", "content": user_input})
    
    # Get response
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=conversation_history
    )
    
    assistant_message = response.choices[0].message.content
    
    # Add assistant response to history
    conversation_history.append({"role": "assistant", "content": assistant_message})
    
    print(f"Assistant: {assistant_message}")

Every request sends the full conversation history — this is how the model maintains context.

Cost implication: Longer conversations cost more because the full history is sent each time. For very long conversations, you may need to truncate older messages.

Error Handling and Best Practices

from openai import OpenAI, RateLimitError, APIError
import time

def safe_api_call(client, messages, model="gpt-4o-mini", retries=3):
    for attempt in range(retries):
        try:
            response = client.chat.completions.create(
                model=model,
                messages=messages,
                timeout=30
            )
            return response
        except RateLimitError:
            if attempt < retries - 1:
                wait_time = 2 ** attempt  # Exponential backoff
                print(f"Rate limit hit. Waiting {wait_time}s...")
                time.sleep(wait_time)
            else:
                raise
        except APIError as e:
            print(f"API error: {e}")
            raise

Best practices:

• Always handle rate limits with exponential backoff
• Set timeout on requests
• Use max_tokens to prevent unexpectedly large (expensive) responses
• Store API keys in environment variables, never in code
• Log usage for cost monitoring

Frequently Asked Questions

How do I get started with the ChatGPT API?

OpenAI account → add payment method → generate API key → install openai library → make your first call. Under 30 minutes to first response.

How much does it cost?

GPT-4o-mini: ~$0.15/1M input tokens. GPT-4o: ~$5/1M input tokens. Most small applications cost pennies per day.

What's the difference between ChatGPT and the API?

ChatGPT is the consumer chat interface. The API gives developers programmatic access to build AI into applications.

What languages can I use?

Python and Node.js have official libraries. Any language that makes HTTP requests can use the API.

What is a system prompt?

A "system" role message that defines the model's behavior for the conversation. Your most powerful tool for customizing AI behavior in your application.

Final Thoughts

An hour of work gets you from zero API knowledge to a working AI-powered application. The fundamentals — API call structure, system prompts, token costs, conversation history — cover 80% of what you need for most applications.

From here: explore OpenAI's documentation for advanced features (function calling, structured outputs, file uploads). And for non-developers who want to get maximum value from ChatGPT without coding, ChatGPT Custom Instructions gives you many of the same customization benefits through the regular interface.