Number of Stars: 26K+ Anthropic’s Official Claude Practical Recipes: Empower Developers to Build AI Applications Efficiently claude-cookbooks is an open-source resource library released by Anthropic that brings together practical Jupyter notebooks and code guides for using Claude AI models. It covers a wide range of scenarios, from basic prompt engineering to advanced agent building, tool integration, and skill customization, providing replicable code snippets and best practices to help developers quickly integrate Claude into their projects.
If you’re just starting out exploring Claude or want to add more power to your AI projects, here are the top 3 notebook picks to check out first.
Notebook 1: RAG (Retrieval-Augmented Generation) Basics Example
Path:https://github.com/anthropics/claude-cookbooks/blob/main/third_party/Wikipedia/wikipedia-search-cookbook.ipynb
Why it’s worth watching
- RAG is one of the core capabilities of modern AI applications (knowledge base Q&A, enterprise document QA, product customer service, etc.) all rely on it).
- This notebook demonstrates in its entirety:
- How to scrape Wikipedia content
- How to segment text, vectorize, and retrieve
- How to answer the retrieved fragment in combination with Claude
After reading it you will get
- The easiest, clearest, and straightforward RAG prototype available
- Intuitive understanding of the working mechanism of “LLM + external knowledge”
- You can replace your own document library with enterprise/course/technical document Q&A
Suitable for: Developers working on knowledge base projects, tool-based applications, and want Claude to “understand your own data”.
Notebook 2: Multimodal: PDF upload + parse + intelligent Q&A
Path:https://github.com/anthropics/claude-cookbooks/blob/main/misc/pdf_upload_summarization.ipynb
Why it’s worth watching
- Claude is very strong in PDF/text structure parsing, and this notebook shows the complete process of actually “eating a document”.
- Content includes:
- How to upload a PDF to Claude
- How to make it automatically extract structures, paragraphs, tables
- How to QA, summarize, and extract highlights based on PDF content
After reading it you will get
- A full template for “AI Auto-Read Documents”
- It can be used for: thesis reading assistant, project contract summary tool, course note generator, meeting minutes analysis
- Directly embed your project to “drop a PDF → automatically understand to AI”
Suitable for: People who often read papers, do document processing, and need automated structured content extraction.
Notebook 3: Image Understanding + Diagram Problem Interpretation (Visual Ability).
Path:https://github.com/anthropics/claude-cookbooks/blob/main/multimodal/reading_charts_graphs_powerpoints.ipynb
Why it’s worth watching
- If you want to truly understand Claude’s “visual modeling” capabilities, this notebook is the most straightforward demonstration.
- Examples include:
- Give Claude an image of a diagram
- Let it do trend analysis, extract data points, and interpret chart meanings
- Combinatorial reasoning of image + text
After reading it you will get
- One set can be used directly to:
- Data graph interpretation AI
- Automatic PPT chart analyzer
- Visual Question Answering (VQA) prototype
- Very good judgment of the boundaries and details of AI’s ability to “read maps”
Suitable for people: people who do visualization, product data analysis, and multi-modal applications. As a visual communication major, you are also very suitable to learn this area.
Why choose these 3?
Claude Cookbooks has a lot of content, but these 3 are the ones that “turn into products the most” :
| Notebook | ability | Actual value to developers |
|---|---|---|
| RAG example | External Knowledge Q&A | Basic features of various AI products |
| PDF parsing | Multimodal + Document Understanding | Automated summarization, knowledge extraction, document assistant |
| Diagram comprehension | Multimodal vision capabilities | Automated chart analysis, visualization enhancement tools |
These three capabilities cover almost 80% of your development needs.
Github：https://github.com/anthropics/claude-cookbooks
Tubing: