Building with the Claude API

This repository contains hands-on exercises and examples from the Anthropic course: Building with the Claude API.

Modules

Module 1: Accessing Claude with the API

📁 001-Accessing-Claude_with-the-API/

Learn the fundamentals of interacting with Claude through the Anthropic API.

Notebook	Description
001_requests.ipynb	Basic API requests, message handling, and multi-turn conversations
001_requests_exercise.ipynb	Build an interactive chatbot using the messages API
002_system_prompt.ipynb	Using system prompts to guide Claude's behavior
002_system_prompt_exercise.ipynb	Create a concise Python code assistant with system prompts
003_temperature.ipynb	Controlling output randomness with the temperature parameter
004_streaming.ipynb	Real-time response streaming and handling stream events
005_controlling_output.ipynb	Message prefilling, stop sequences, and structured output (JSON)

Module 2: Prompt Evaluation

📁 002-Prompt-evaluation/

Learn how to systematically evaluate and grade prompt outputs using datasets, model-based grading, and code-based validation.

Notebook	Description
001_generating_dataset.ipynb	Generate evaluation datasets for testing prompts
002_running_eval.ipynb	Run evaluations across test cases and collect results
003_model_based_grading.ipynb	Use Claude to grade and evaluate AI-generated solutions
004_code_based_grading.ipynb	Combine syntax validation (JSON, Python, Regex) with model grading
005_exercise.ipynb	Exercise: Enhance the model grader with solution criteria

Module 3: Prompt Engineering Techniques

📁 003-Prompt-engineering/

Master essential prompt engineering techniques to improve Claude's output quality, consistency, and accuracy.

Notebook	Description
001_prompting.ipynb	Introduction to prompt engineering with the PromptEvaluator framework
002_being_clear_direct.ipynb	Writing clear and direct prompts for better results
003_being_specific.ipynb	Using guidelines and step-by-step instructions for complex tasks
004_xml.ipynb	Structuring prompts with XML tags for clarity and organization
005_providing_examples.ipynb	One-shot and multi-shot examples to guide model behavior
006_prompting_exercise.ipynb	Exercise: Extract topics from scholarly articles

Module 4: Tool Use with Claude

📁 004-Tool-use-with-Claude/

Learn how to extend Claude's capabilities by defining tools, handling tool calls, and building multi-turn agentic workflows.

Notebook	Description
001_tools_function.ipynb	Introduction to tool functions for accessing real-time information
002_tools_schemas.ipynb	Creating JSON schemas to define tool arguments
003_tools_handling_message_blocks.ipynb	Handling multi-block messages with text and tool use content
004_tools_sending_tool_results.ipynb	Executing functions and sending results back to Claude
005_tools_multi-turn-conversations.ipynb	Calling multiple tools in sequence to answer a single question
006_tools_implementing_multiple_turns.ipynb	Building a conversation loop that runs until tool use completes
007_tools_using_multiple_tools.ipynb	Integrating multiple tools into a single implementation
008_tools_batch_tools.ipynb	Running multiple tool calls in parallel within a single message
009_tools_structured_data.ipynb	Using tools to extract structured data reliably
010_tools_fine_grained_tool_calling.ipynb	Combining tool use with streaming for real-time updates
011_tools_text_edit.ipynb	Using Claude's built-in text editor tool for file operations
012_tools_web_search.ipynb	Enabling Claude's built-in web search tool for current information

Module 5: Retrieval-Augmented Generation (RAG)

📁 005-Retrieval-Augmented-Generation/

Build RAG pipelines from scratch, covering chunking strategies, embeddings, vector search, lexical search, and advanced retrieval techniques.

Additional setup required for this module:

• Install the Voyage AI package: pip install voyageai
• Add your Voyage AI API key to .env: VOYAGE_API_KEY="your_key_here" (Get one here)

Notebook	Description
001_introducing-RAG.ipynb	Text chunking strategies: size-based, structure-based, and semantic-based
002_embeddings.ipynb	Generating text embeddings for semantic search over document chunks
003_vectordb.ipynb	Complete RAG flow with chunking, embeddings, vector storage, and similarity search
004_bm25.ipynb	BM25 lexical search as a complement to semantic search
005_hybrid.ipynb	Hybrid search combining semantic and lexical retrieval with Reciprocal Rank Fusion
006_reranking.ipynb	Reranking retrieved documents with Claude for improved accuracy
007_contextual.ipynb	Contextual retrieval to preserve document context in chunked passages

Module 6: Features of Claude

📁 006-Feature-of-Claude/

Explore Claude's advanced built-in features including extended thinking, vision, PDF processing, citations, caching, and code execution.

Notebook	Description
001_thinking.ipynb	Extended thinking for complex reasoning before generating responses
002_images.ipynb	Vision capabilities for analyzing and understanding images
003_pdf.ipynb	Reading and analyzing PDF files for document processing
004_citations.ipynb	Citations to trace answers back to specific source locations
005_caching.ipynb	Prompt caching to speed up responses and reduce costs
006_code_execution.ipynb	Files API and code execution for delegating complex tasks

Module 7: Model Context Protocol (MCP)

📁 007-Model-Context-Protocol/

Learn how to build MCP servers and clients to extend Claude with tools, resources, and reusable prompts through a standardized protocol.

Additional setup required for this module:

• Install uv (recommended): curl -LsSf https://astral.sh/uv/install.sh | sh
• Run uv sync inside the cli_project directory to install dependencies
• Add CLAUDE_MODEL to the project's .env file (e.g., CLAUDE_MODEL="claude-sonnet-4-5")

File	Description
NOTES.md	Concise reference notes on MCP concepts and the sample project
cli_project/mcp_server.py	MCP server defining tools, resources, and prompts using FastMCP
cli_project/mcp_client.py	MCP client connecting to servers via stdio transport
cli_project/main.py	Entry point wiring the MCP client with an interactive CLI chat
cli_project/core/chat.py	Multi-turn tool execution loop with Claude
cli_project/core/cli_chat.py	CLI chat with @ document mentions and / command prompts
cli_project/core/tools.py	Tool discovery and execution across multiple MCP clients
cli_project/core/cli.py	Interactive CLI with tab-completion and auto-suggest

Module 8: Anthropic Apps

📁 008-Anthropic-Apps/

Explore two powerful Anthropic applications — Claude Code and Computer Use — and build a real MCP document tools server using the included app_starter project.

Additional setup required for this module (app_starter/):

• Install uv if not already installed: curl -LsSf https://astral.sh/uv/install.sh | sh
• cd 008-Anthropic-Apps/app_starter
• uv venv && source .venv/bin/activate
• uv pip install -e .
• Run the MCP server: uv run main.py
• Run tests: uv run pytest

File	Description
NOTES.md	Notes on Claude Code, Computer Use, /init, and MCP server registration
app_starter/main.py	FastMCP server entry point registering document and math tools
app_starter/tools/document.py	Tool to convert PDF/DOCX files to Markdown via MarkItDown
app_starter/tools/math.py	Reference tool implementation demonstrating the expected tool definition pattern
app_starter/tests/test_document.py	pytest tests for the document conversion tool

Module 9: Agents and Workflows

📁 009-Agents-and-workflows/

Learn the core strategies for building multi-step AI systems — chaining, routing, parallelization workflows — and fully autonomous agents powered by tools.

File	Description
NOTES.md	Concise reference notes on workflows, agents, and when to use each strategy

Setup

Prerequisites

• Python 3.8+
• An Anthropic API key (Get one here)

Installation

1. Clone the repository:

   git clone https://github.com/jaygaha/Building-with-the-Claude-API.git
   cd Building-with-the-Claude-API

2. Create and activate a virtual environment:

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configure your API key:

   cp .env-example .env

   Then edit .env and add your Anthropic API key.

5. Launch Jupyter:

   jupyter notebook

Resources

• Anthropic API Documentation
• Claude API Reference
• Anthropic Courses