What Python Import Statements Actually Do (Beginner's Guide)

The first few lines of almost every AI Python script look like this:

import os
import anthropic
from dotenv import load_dotenv

These lines aren't doing anything yet, they're loading tools. Once you can read them, you'll know exactly what a script depends on before running a single line of logic.

Key Takeaways

• import x loads a whole library, you access its tools with x.something.
• from x import y pulls one specific thing out, so you use y directly, no prefix.
• import x as y gives a library a shorter nickname, common in data science code.
• Standard library imports (os, json) come with Python, third-party imports (anthropic, dotenv) need pip install first.
• Recognizing which type an import is tells you what happens if that line, or the install behind it, is missing.

What a library actually is

The same way you learned to trace a file operation line by line, you can learn to trace an import line by line, once you know the pattern, you'll never see it as a mystery block again.

A library, also called a package or module, is someone else's code you can use in your own program. Instead of building every tool from scratch, you import tools that are already made and tested. import anthropic says: go find the anthropic library, which I installed earlier, and load all its tools so I can use them.

The import runs once, at the top of the script. After that, every tool the library provides is available for the rest of the file. You never have to re-import it mid-script.

The three import patterns

You'll only ever see three shapes. Once you can name them, you can read any imports block.

Pattern 1, import the whole library

import anthropic

Access its tools with dot notation: anthropic.Anthropic(), anthropic.APIError. Use this when you need multiple things from the library.

Pattern 2, import one specific thing

from anthropic import Anthropic

Access it directly: Anthropic(), no anthropic. prefix needed. Use this when you only need one class or function and don't want to type the full path every time.

Pattern 3, import with an alias

import numpy as np

Access it as np.array(...) instead of numpy.array(...). Use this when a library's name is long and you'll type it constantly. Less common in simple AI scripts, but you'll see it often in data science code.

The three types of libraries

Every import falls into one of three buckets, and the bucket tells you what breaks if the import fails.

Standard library comes with Python, no install needed. import os gives you operating system functions like file paths and environment variables. import json parses and generates JSON. These never need pip install, they're always available.

Third-party packages are installed via pip. import anthropic needs pip install anthropic first. from dotenv import load_dotenv needs pip install python-dotenv. If you haven't installed it, the import fails with ModuleNotFoundError, that's your signal the pip install step was skipped.

Local files are your own code in the same project. from utils import build_prompt imports from a file you wrote called utils.py. No install needed, Python finds it because it's sitting in the same folder.

Decode a real imports block

import os # standard library, access environment variables
import json # standard library, parse API responses
import anthropic # third-party, the Claude SDK
from dotenv import load_dotenv # third-party, load .env files

Two standard, two third-party. This exact block, or something close to it, sits at the top of most beginner AI scripts. I've seen this four-line pattern open more Claude scripts than almost any other combination, it's worth being able to read on sight.

Match the Import

Six real import lines. For each, work out the type and what removing it would break before checking the answer.

import os

import anthropic

from dotenv import load_dotenv

import json

from tools import web_search

import datetime

The imports block is the ingredient list. Once you can read it, you know exactly what a script depends on, and what breaks, before running a single line of logic. That's the same skill you'll keep building through the rest of the Launchpad path.

Your Task