Claude API Basics - Keys, a First Request, Streaming and Errors
Make your first call to the Claude API safely: get a key, keep it out of your code, send a request, stream the reply and handle errors.
The Claude API lets your own programs talk to Claude, so you can summarize tickets, draft replies or power an assistant inside your own tools. Getting a first response is quick. Doing it safely, with the key kept secret and failures handled calmly, is what separates a demo from something you can leave running.
Related guides
Getting Reliable JSON and Structured Output from AI
Techniques for making a model return clean JSON every time, from prompt wording and schemas to validation and retry loops in your code.
6 min read
Building Your First AI Agent, Step by Step
Understand what an agent is, then build a simple one: a goal, a loop, a few tools and clear stopping rules, with a framework-free Python sketch.
7 min read
Getting Started with Claude: A Beginner's Walkthrough
How to start using Claude from Anthropic, what its main features are for, and a few first tasks to try, with tips for getting the best answers.
5 min read