Build an AI-Powered App with Claude Course

Lesson 1 • Making Your First API Call

Walks through a minimal request-response cycle using the Claude API. Confirms environment setup and introduces core request parameters hands-on.

Lesson 2 • Setting Up Your Development Environment

Guides installation of required tools, API key management, and project scaffolding. Ensures every student has a working environment before writing API calls.

Lesson 3 • Claude's Design and Capabilities

Examines Claude's training approach, safety alignment, and supported modalities. Clarifies what Claude can and cannot do before any code is written.

Lesson 1 • Prompt Testing and Iteration

Introduces systematic evaluation of prompt variants using test cases. Establishes a feedback loop that students apply throughout the entire course.

Lesson 2 • System Prompts and User Turns

Distinguishes system-level instructions from user messages in the messages array. Shows how layering these controls Claude's behaviour across a session.

Lesson 3 • Anatomy of an Effective Prompt

Breaks down role, instruction, context, and output-format components. Provides a reusable template structure applicable to all subsequent projects.

Lesson 4 • Few-Shot and Chain-of-Thought Techniques

Teaches example-based prompting and step-by-step reasoning elicitation. Improves accuracy on complex tasks introduced in later applied chapters.

Lesson 1 • Error Handling and Retry Logic

Maps API error codes to causes and implements exponential backoff. Produces resilient apps that recover gracefully from rate limits and transient failures.

Lesson 2 • Model Selection and Versioning

Compares available Claude model tiers by speed, cost, and capability. Teaches version pinning to prevent unexpected behaviour changes in deployed apps.

Lesson 3 • Token Counting and Cost Management

Uses the token-counting endpoint and prompt caching to control spend. Directly supports budget planning covered in the advanced optimisation chapter.

Lesson 4 • Streaming Responses in Real Time

Implements server-sent events to stream Claude tokens to the client. Enables responsive UIs that display output progressively rather than after full generation.

Lesson 1 • Conversation State and Memory

Explains how the messages array maintains context across turns. Establishes the state management pattern all subsequent chat features depend on.

Lesson 2 • Persisting Conversation History

Stores and retrieves conversation threads using a database. Enables users to resume sessions, a prerequisite for the multi-user app chapter.

Lesson 3 • Designing Chat UI Components

Builds message bubbles, input fields, and streaming indicators using a frontend framework. Connects visual components to the API layer established earlier.

Lesson 4 • Persona and Tone Customisation

Uses system prompts to give the chatbot a consistent brand voice and persona. Demonstrates how prompt engineering from Chapter 2 applies to product design.

Lesson 1 • Executing Tool Calls Server-Side

Parses Claude's tool_use blocks and dispatches to real functions or APIs. Closes the loop by returning tool results back to Claude for final response.

Lesson 2 • Building Practical Tool Integrations

Implements web search, database query, and calculation tools as concrete examples. Gives students reusable patterns for any external service integration.

Lesson 3 • Tool Safety and Input Validation

Validates and sanitises all tool inputs before execution to prevent injection attacks. Applies security principles that carry forward into the deployment chapter.

Lesson 4 • Defining Tools with JSON Schema

Writes tool definitions using JSON Schema to specify parameters and types. Accurate schemas reduce parsing errors in tool call responses.

Lesson 1 • Retrieval-Augmented Generation Basics

Embeds document chunks and retrieves relevant passages before prompting Claude. Grounds Claude's answers in source documents, reducing hallucination on factual queries.

Lesson 2 • Building a Document Q&A Feature

Combines upload, extraction, retrieval, and Claude prompting into an end-to-end feature. Integrates all multimodal skills from this chapter into one deployable component.

Lesson 3 • Sending Images to Claude

Encodes images as base64 or URL references in the messages array. Unlocks visual understanding features built upon in the document pipeline section.

Lesson 4 • Processing PDF and Text Documents

Extracts and chunks text from PDFs for inclusion in Claude prompts. Handles documents that exceed context limits using the chunking strategies introduced here.

Lesson 1 • Caching Strategies for AI Responses

Applies semantic and exact-match caching to reduce redundant API calls. Lowers latency and cost for repeated or similar queries in production traffic.

Lesson 2 • Rate Limiting and Quota Management

Enforces per-user rate limits and tracks API quota consumption in real time. Protects both cost budgets and service availability under high user load.

Lesson 3 • Security and API Key Protection

Implements server-side key storage, request authentication, and CORS policies. Prevents credential exposure vulnerabilities common in early AI app prototypes.

Lesson 4 • Logging, Monitoring, and Observability

Instruments request traces, latency metrics, and error rates for Claude API calls. Enables rapid diagnosis of production issues using structured log data.

Lesson 1 • Cost Optimisation Techniques

Reduces token spend through prompt compression, caching, and model routing. Directly applies token-counting skills from Chapter 3 to real cost reduction.

Lesson 2 • Output Quality Improvement Loops

Uses evaluation results to drive iterative prompt and architecture changes. Closes the build-measure-improve cycle for sustained quality gains over time.

Lesson 3 • Designing an Evaluation Framework

Defines metrics, test datasets, and scoring rubrics for Claude output quality. Provides the measurement foundation all optimisation work in this chapter relies on.

Lesson 4 • A/B Testing AI Features

Runs controlled experiments comparing prompt variants or model versions in production. Provides statistical confidence before committing to any optimisation change.