The Claude Developer Platform

Knowing the pieces (Console, Messages API, SDKs, Agent SDK, Managed Agents) tells you where to start and which docs to read.

Claude is the family of language models made by Anthropic. You can chat with Claude in the claude.ai app, but to use Claude inside your own program you call it through the Claude API, part of what Anthropic calls the Claude Developer Platform.

Think of it like ordering from a kitchen through a service window: your program sends a written order (a request), and Claude sends back a reply (a response). Almost everything goes through one web address (an endpoint), the Messages API: you send a list of messages, Claude answers with the next message.

The pieces you will meet:

Claude Console (platform.claude.com): the website where you create API keys, see usage and set limits.
Client SDKs: official libraries that wrap the HTTP calls. In Python the package is anthropic.
Docs: the developer docs now live under platform.claude.com/docs (older links on docs.anthropic.com and docs.claude.com redirect or are outdated).
Claude Agent SDK: a separate library that gives your program the same agent loop and tools that power Claude Code.
Claude Managed Agents: a hosted option where Anthropic runs the agent for you (in beta in late 2026).

This course uses the Messages API through the Python SDK, because every other piece builds on it. Once you can send messages, handle tools and control cost there, the higher-level options are easy to judge.

One thing to expect: the platform changes fast. Model names, beta features and limits change every few months, so always check the docs for the current values instead of copying old blog posts.

Example

# The whole platform in one picture (pseudo-code, not runnable):
#
#   your Python program
#        |  request: model + messages   (sent by the `anthropic` SDK)
#        v
#   Messages API  (https://api.anthropic.com/v1/messages)
#        |  response: Claude's reply + token usage
#        v
#   your Python program
#
# Higher-level options built on top of the same models:
#   Claude Agent SDK  -> agent loop + built-in tools inside your app
#   Managed Agents    -> Anthropic hosts and runs the agent

Remember

Your code talks to Claude through the Messages API
The Python SDK is the anthropic package; keys live in the Claude Console
Current docs are at platform.claude.com/docs; details change often

Watch out for

Following an old tutorial for model names or beta flags; check the current docs first.

Check yourself

Which part of the platform do you use to create an API key?

Official docs: platform.claude.com/docs/en/intro

Example

Remember

Watch out for

Check yourself

Builds on

Unlocks