LLM Provider SDKs: OpenAI & Anthropic

Introduction

OpenAI and Anthropic are the two most widely used LLM provider SDKs, and nearly every orchestration framework later in this course supports both as a backend. Learning their two APIs directly, before adding any framework on top, makes it much easier to understand what a framework is actually doing for you.

A Real-Life Analogy First

Think of OpenAI and Anthropic as two different, excellent translation agencies. Both employ highly skilled translators (the models), both accept a document to translate (your messages) and hand back a translated document (the response) — but each agency has its own intake form: one asks you to note special instructions in a dedicated field at the top (Anthropic's separate system parameter), while the other lets you slip that same instruction in as the first item on the document itself (OpenAI's system message inside the list). Once you have filled out one agency's form, the second one's form looks instantly familiar.

OpenAI: The GPT SDK

Use case: the openai package calls OpenAI's GPT models through a chat-completions style API — a list of role-tagged messages in, a generated message out. It also supports streaming, structured outputs, and tool/function calling.

pip install openai

OpenAI in Action

from openai import OpenAI

client = OpenAI()  # reads OPENAI_API_KEY from the environment

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {"role": "system", "content": "You are a concise coding assistant."},
        {"role": "user", "content": "What does the yield keyword do in Python?"},
    ],
    temperature=0.3,
)

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

Anthropic: The Claude SDK

Use case: the anthropic package calls Anthropic's Claude models through a Messages API. It looks almost identical to OpenAI's shape by design, but keeps the system prompt as a separate top-level parameter rather than a message in the list.

pip install anthropic

Anthropic in Action

import anthropic

client = anthropic.Anthropic()  # reads ANTHROPIC_API_KEY from the environment

response = client.messages.create(
    model="claude-sonnet-5",
    max_tokens=200,
    system="You are a concise coding assistant.",
    messages=[
        {"role": "user", "content": "What does the yield keyword do in Python?"}
    ],
)

print(response.content[0].text)

OpenAI vs Anthropic: Key Differences

In practice, most orchestration frameworks (LangChain, LlamaIndex — covered in lessons 9 and 10) abstract these differences behind a shared interface, so switching providers often means changing one line of configuration rather than rewriting your application logic.

Where You've Already Used Both

Even if you have never written a line of code calling either SDK, you have very likely used products built on top of both.

Common Mistakes

Best Practices

Frequently Asked Questions

Key Takeaways

Summary

OpenAI and Anthropic's SDKs are the entry point to nearly everything else in this course — every framework and pipeline ahead is ultimately built on top of a call that looks like the two shown here.