← All open source projects

Claude Cookbooks

anthropics/claude-cookbooks

Claude Cookbooks is a set of practical notebooks and recipes for working with the Claude API.

Forks 5,411
Author anthropics
Language Jupyter Notebook
License MIT
Synced 2026-06-27

What it is

Claude Cookbooks is a collection of practical recipes for working with Claude. Instead of only listing parameters, the examples show whole scenarios: prepare data, call a model, process output, and check quality.

The project helps developers who understand LLM basics but want applied patterns in code and notebooks.

What is inside

The repository contains Jupyter notebooks, Python examples, and material for information extraction, document handling, tools, evaluation, assistants, and integrations.

The cookbook format is useful because an example can be run, inspected cell by cell, and adapted.

How it is used

Developers take a relevant recipe, replace data, keys, and model choices, then observe how the result changes.

A product system cannot copy a recipe blindly. Error handling, limits, data safety, logging, and quality checks must be added.

Strengths and limits

The strength is closeness to real code. The repository shows the surrounding task, not only API parameters.

The limitation is that recipes are not full architecture. They explain a technique; teams still design storage, access, and risk controls.

The best use is a starting lab, followed by moving stable parts into normal service code.

The practical value of Claude Cookbooks is easiest to see through a small verifiable scenario: take the task the project was made for and follow it to a result. Claude Cookbooks shows practical Claude usage: API calls, document handling, evaluation, tools, and applied examples. That separates real usefulness from a nice description.

If Claude Cookbooks stays in use beyond the first experiment, maintenance starts to matter as much as features: updates, clear responsibility boundaries, testable examples, and the project’s place in the existing system. That is where real strengths and limits usually appear.

Example

Контур рецепта с моделью

Пример показывает общий вид прикладного ноутбука: подготовить вход, вызвать модель и отдельно обработать ответ.

Language: Python
from anthropic import Anthropic

client = Anthropic()
message = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=300,
    messages=[{"role": "user", "content": "Summarize this document"}],
)

print(message.content[0].text)