Claude API Fundamentals

With the Claude API (Claude Developer Platform) you can build Claude into your own app. Everything goes through one main endpoint: POST /v1/messages (the Messages API). Tools, thinking and structured output are all features of this same endpoint.

5.1 Setup

1. Create an account on platform.claude.com (Console) and add billing.
2. Create an API key (inside a workspace). Do not hardcode the key in code — keep it in the ANTHROPIC_API_KEY environment variable.
3. Install the SDK: Python pip install anthropic, TypeScript npm install @anthropic-ai/sdk, C# dotnet add package Anthropic. There are also official SDKs for Java, Go, Ruby and PHP.

5.2 First request (Python, C#, cURL)

import anthropic
client = anthropic.Anthropic()          # reads ANTHROPIC_API_KEY from the environment

response = client.messages.create(
    model="claude-opus-5-5",
    max_tokens=16000,
    system="You are a helpful assistant that answers in Hinglish.",
    messages=[{"role": "user", "content": "What is an index in SQL?"}],
)
for block in response.content:          # content is a list (text, thinking, tool_use...)
    if block.type == "text":
        print(block.text)
print(response.stop_reason, response.usage.input_tokens, response.usage.output_tokens)

using Anthropic;
using Anthropic.Models.Messages;

AnthropicClient client = new();     // reads ANTHROPIC_API_KEY from the environment

var message = await client.Messages.Create(new MessageCreateParams
{
    Model = "claude-opus-5-5",
    MaxTokens = 1024,
    Messages = [ new UserMessageParam { Content = "Hello, Claude" } ],
});
Console.WriteLine(message);

curl https://api.anthropic.com/v1/messages \
  -H "content-type: application/json" \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2026-10-21" \
  -d '{"model": "claude-opus-5-5", "max_tokens": 1024,
       "messages": [{"role": "user", "content": "Hello, Claude"}]}'

5.3 Request parameters

5.4 Response and stop_reason

{
  "id": "msg_01...",
  "type": "message",
  "role": "assistant",
  "model": "claude-opus-5-5",
  "content": [{"type": "text", "text": "An index is a data structure..."}],
  "stop_reason": "end_turn",
  "usage": {"input_tokens": 25, "output_tokens": 180,
            "cache_read_input_tokens": 0, "cache_creation_input_tokens": 0}
}

5.5 Multi-turn conversation

history = []
def chat(user_text):
    history.append({"role": "user", "content": user_text})
    r = client.messages.create(model="claude-opus-5-5", max_tokens=4000, messages=history)
    history.append({"role": "assistant", "content": r.content})   # append the full content back
    return "".join(b.text for b in r.content if b.type == "text")

chat("My name is Rahul")
print(chat("What is my name?"))     # remembered because of the history

5.6 Streaming

In streaming, the response arrives in pieces over Server-Sent Events (SSE), so the user sees text immediately and long requests do not time out. Event order: message_start → content_block_start → content_block_delta (repeatedly) → content_block_stop → message_delta (stop_reason, usage) → message_stop. A ping can also arrive in between.

with client.messages.stream(
    model="claude-opus-5-5", max_tokens=64000,
    messages=[{"role": "user", "content": "Write a long article on microservices"}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)
    final = stream.get_final_message()     # the complete message object

5.7 Errors and rate limits

• There are three kinds of rate limits: RPM (requests/min), ITPM (input tokens/min), OTPM (output tokens/min). Limits grow as your usage tier goes up.
• The SDK retries 2 times by default (408, 409, 429, 5xx) with exponential backoff. Default timeout is 10 minutes.
• In error handling, catch from specific to general: RateLimitError → APIStatusError → APIConnectionError.