Claude Cookbooks
A set of Jupyter notebook recipes showing how to build with the Claude API.
- Stars
- 54.1k
- Forks
- 6.5k
- Licence
- MIT
- Language
- Jupyter Notebook
- Last update
- 7 days ago
- Type
- Learning resource
What it is
This repository is a collection of Jupyter notebooks and guides with copy-able code examples for using Claude, Anthropic's AI model, through its API. Each recipe covers a specific task, such as classification, retrieval augmented generation, tool use, vision, summarization, prompt caching, or automated evaluations. Developers use it as a starting point and adapt the snippets into their own projects.
- Build a customer service agent that uses tools, based on the tool use recipe
- Create a retrieval augmented generation pipeline with a vector database such as Pinecone
- Extract text or data from images, charts, or forms using vision examples
- Summarize PDFs by parsing them and passing the text to Claude
- Set up prompt caching or a cost optimization checklist for an agent
Great for
- Python developers who want working examples of Claude API calls
- Developers exploring tool use or retrieval augmented generation with Claude
- Teams evaluating how to add vision or PDF handling to an application
- People new to the Claude API who want to follow guided examples
Not ideal for
- Non-programmers, since the recipes are code notebooks that assume coding knowledge
- People looking for a finished end-user app, because this is example code rather than a product
- Anyone needing non-Python examples, since the README says the code is primarily Python and other languages must be adapted
How to get it
- 1
Clone and build from source
You'll need: Python · A Claude API key · Jupyter Notebook or JupyterLab
Clone the repository
git clone https://github.com/anthropics/claude-cookbooks - 2
Run it
- Sign up for a Claude API key and store it securely
- Clone the repository and install Python and Jupyter
- Open the notebook for the recipe that matches your task
- Run the cells in order, supplying your API key as the notebook expects
- 3
Get the most out of it
- Start with the Claude API Fundamentals course linked in the README if you are new to the API
- Combine the tool use and retrieval recipes to build an agent that answers from your own data
- Use the prompt caching recipe to reduce cost on repeated long prompts
- Follow the automated evaluations recipe to measure prompt changes systematically
Downloads
Strengths
- Covers many common tasks, from classification to vision to agents, in one place
- Code is meant to be copied and integrated into real projects
- Released under the MIT license, which permits reuse
Trade-offs
- The README is short and does not list versions or detailed setup steps, so you must read each notebook for specifics
- Some recipes link to external repositories or third-party services that need their own accounts or setup
- Examples may reflect the Claude API at the time they were written, so check current docs before relying on them
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| claude-cookbooks (this) | 54.1k | Learning resource | intermediate | 7 days ago | Python developers who want working examples of Claude API calls |
| langchain | 149k | Framework | intermediate | 4 days ago | Python developers building LLM-powered applications |
| llama_index | 52.8k | Library | intermediate | 10 days ago | Python developers building retrieval-augmented generation (RAG) apps over private documents |
| openai-cookbook | 76.9k | Learning resource | beginner | 5 days ago | Developers who are new to the OpenAI API and want working examples |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.