The Claude API is one of the most developer-friendly AI APIs available in 2025. The official SDKs for Python and TypeScript are well-documented, the pricing is transparent, and the API itself is remarkably consistent across model versions. This guide walks through the complete setup process and the patterns you will actually use in production.
Setting Up Your Account and API Key
Start at console.anthropic.com. Create an account and navigate to Settings, then API Keys. Click Create Key, give it a descriptive name like your-app-dev, and copy the key immediately — it begins with sk-ant- and is shown only once. Store it securely: never hardcode it in source code. The correct approach is an environment variable: export ANTHROPIC_API_KEY=your-key-here, or use a .env file excluded from git.
Before making API calls, add a payment method under Billing. Anthropic offers free credits for new accounts, but setup is required to access higher rate limits. Set a monthly spending limit here to prevent surprise bills — Anthropic will pause access rather than overcharging.
Installing the SDK
Anthropic maintains official SDKs for Python and TypeScript. Install the Python SDK with pip install anthropic, or the TypeScript SDK with npm install @anthropic-ai/sdk. Both use the same API under the hood — choose based on your existing stack.
Once installed, the SDK reads your ANTHROPIC_API_KEY environment variable automatically. No explicit configuration needed for basic use.
Your First API Call
The most basic working example in Python:
import anthropic
client = anthropic.Anthropic()
message = client.messages.create(
model=claude-sonnet-4-20250514,
max_tokens=1024,
messages=[{role: user, content: What is the capital of France?}]
)
print(message.content[0].text)
The key parameters are model (use dated model IDs like claude-sonnet-4-20250514 in production for reproducibility — claud-sonnet-4-latest is fine for development), max_tokens (controls maximum response length, set high enough to avoid truncation), and messages (a list of role/content pairs). For production use, always set an appropriate max_tokens for your expected output length.
System Prompts and Multi-Turn Conversations
Claude uses a separate system parameter for persistent instructions, distinct from messages in the conversation array. This is where you set behavior guidelines that persist across the entire conversation: client.messages.create(…, system=You are a senior Python developer. Be concise., messages=[…]).
For multi-turn conversations, append both user and assistant turns to the messages array. Claude maintains context across all messages — you do not resend the full conversation history, just add new turns.
Streaming Responses
Streaming is essential for chat interfaces where you want to display output as it generates rather than waiting for the complete response. In Python:
with client.messages.stream(
model=claude-sonnet-4-20250514,
max_tokens=1024,
messages=[{role: user, content: Tell me a joke.}]
) as stream:
for text in stream.text_stream:
print(text, end=, flush=True)
This yields tokens as they generate, which you can push to a WebSocket, SSE endpoint, or directly to a frontend chat interface.
Tool Use (Function Calling)
One of Claude most powerful features is tool use — the ability to define functions that Claude can call during a conversation. Define tools with a JSON schema:
tools = [{name: get_weather, description: Get current weather for a city, input_schema: {type: object, properties: {city: {type: string}}, required: [city]}}]
When you send a message and Claude determines it needs to call a tool, the API response will include a tool_use content block rather than a text response. Your code executes the function and sends the result back, and Claude incorporates it into its next response. This is the foundation of agentic AI workflows.
Choosing the Right Model
Claude Sonnet 4.6 offers the best balance of quality and cost for most production applications at $3/input and $15/output per million tokens. Use Claude Opus 4.8 for complex reasoning tasks where quality matters more than cost. Use Claude Haiku for high-volume, simple classification or extraction tasks where speed and cost efficiency are priorities.
All Claude models share a 200k token context window — roughly 150,000 words or a full novel — making them uniquely suited for long-document tasks that would require chunking or retrieval augmentation with other models.