Claude API & SDKs

📖 5 min read claudeanthropicapisdkdevelopmentprovider-api Key Info

Complete guide to the Claude API — Messages API, tool use, streaming, structured outputs, prompt caching, batch processing, web search, computer use, SDKs, and cost optimization.

Key Takeaways

• Messages API is the primary endpoint: send text + images, receive text responses with tool use, streaming, and structured outputs
• Prompt caching: 10% of input price for cache reads (90% savings). Batch API: 50% discount for async. Combine for up to 95% savings
• All SDKs (Python, TypeScript, Java, Go) support streaming, tool use, and the full feature set
• Web search ($10/1K searches), web fetch (free), computer use, and code execution tools available

On this page 18

OverviewGetting StartedMessages APIContent Blocks (Multimodal Input)Tool Use (Function Calling)StreamingStructured OutputsPrompt Caching — 90% Cost SavingsBatch Processing — 50% DiscountBuilt-in ToolsWeb Search — $10/1K SearchesWeb Fetch — Free (Tokens Only)Computer UseCode ExecutionCost Optimization StrategiesRate LimitsSDKs — Quick ReferenceWhere Next

Getting Started

# Get your API key from https://console.anthropic.com
export ANTHROPIC_API_KEY="your-api-key"

# Python SDK
pip install anthropic

import anthropic
client = anthropic.Anthropic()

message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello, Claude!"}]
)
print(message.content[0].text)

// TypeScript SDK
npm install @anthropic-ai/sdk

import Anthropic from '@anthropic-ai/sdk';
const client = new Anthropic();

const message = await client.messages.create({
  model: 'claude-sonnet-4-6',
  max_tokens: 1024,
  messages: [{ role: 'user', content: 'Hello, Claude!' }],
});
console.log(message.content[0].text);

Messages API

The Messages API is the primary endpoint. Key capabilities:

Content Blocks (Multimodal Input)

message = client.messages.create(
    model="claude-opus-4-8",
    max_tokens=1024,
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "Describe this image:"},
            {"type": "image", "source": {
                "type": "base64",
                "media_type": "image/png",
                "data": base64_image_data
            }}
        ]
    }]
)

Tool Use (Function Calling)

tools = [{
    "name": "get_weather",
    "description": "Get current weather for a location",
    "input_schema": {
        "type": "object",
        "properties": {
            "location": {"type": "string", "description": "City and state"}
        },
        "required": ["location"]
    }
}]

message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    tools=tools,
    messages=[{"role": "user", "content": "What's the weather in San Francisco?"}]
)

Streaming

with client.messages.stream(
    model="claude-haiku-4-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Write a haiku about AI"}]
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

Structured Outputs

message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Extract names from: Alice met Bob at 3 PM"}],
    response_format={
        "type": "json_schema",
        "json_schema": {
            "name": "people_extraction",
            "schema": {
                "type": "object",
                "properties": {
                    "people": {"type": "array", "items": {"type": "string"}}
                }
            }
        }
    }
)

Prompt Caching — 90% Cost Savings

Prompt caching stores frequently-used portions of your prompt so subsequent requests read from cache at 10% of the base input price.

Operation	Multiplier vs Base Input	Duration
5-min cache write	1.25×	5 minutes
1-hour cache write	2×	1 hour
Cache read (hit)	0.1×	Same as the write

import anthropic
client = anthropic.Anthropic()

# Write to cache (1.25x for 5-min, 2x for 1-hour)
message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    system=[
        {
            "type": "text",
            "text": "You are a financial analyst...",  # long system prompt
            "cache_control": {"type": "ephemeral"}
        }
    ],
    messages=[{"role": "user", "content": "Analyze Q1 results"}]
)

When to use:

• Long system prompts (instructions, role definitions)
• Repeated document analysis (same PDF, different questions)
• Conversation threads with shared context
• Multi-turn agent interactions

ROI: Cache pays off after just 1 read for 5-min writes, or 2 reads for 1-hour writes.

Batch Processing — 50% Discount

For async, high-volume workloads:

message = client.messages.beta.batches.create(
    model="claude-haiku-4-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Classify: ${text}"}],
    # Results available within 24 hours, typically much faster
)

Model	Batch Input	Batch Output	vs Standard
Opus 4.8	$2.50 / 1M	$12.50 / 1M	50% off
Sonnet 4.6	$1.50 / 1M	$7.50 / 1M	50% off
Haiku 4.5	$0.50 / 1M	$2.50 / 1M	50% off

Built-in Tools

Web Search — $10/1K Searches

message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    tools=[{"type": "web_search_20260209"}],
    messages=[{"role": "user", "content": "Latest GPT-5 pricing compared to Claude?"}]
)

Charged per search executed, plus standard token costs for search results as input.

Web Fetch — Free (Tokens Only)

message = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    tools=[{"type": "web_fetch_20260209"}],
    messages=[{"role": "user", "content": "Fetch and summarize the content at https://example.com"}]
)

No additional charge beyond standard token costs. Use max_content_tokens to cap fetched content size.

Computer Use

# Requires Claude computer use setup (Docker, etc.)
# Full setup: https://docs.anthropic.com/en/docs/agents-and-tools/computer-use

Adds 735 input tokens for tool definition + screenshot images as vision tokens.

Code Execution

Free when combined with web search or web fetch. Otherwise: $0.05/hour/container (1,550 free hours/month).

Cost Optimization Strategies

Strategy	Savings	When to Apply
Prompt Caching	Up to 90% on input	Repeated system prompts, same-document analysis
Batch API	50% on token costs	Non-urgent, high-volume processing
Model Routing	30-60%	Route simple queries to Haiku, complex to Opus
Prompt Caching + Batch	Up to 95%	High-volume repeated analysis (nightly reports, classification)
Streaming	Better UX (earlier TTFT)	Interactive applications

Rate Limits

Tier	RPM (Opus/Sonnet)	RPM (Haiku)	TPM (Tokens/min)
Tier 1 (Free/Starter)	50	100	100K
Tier 2 (Build)	1,000	2,000	2M
Tier 3 (Scale)	5,000	10,000	10M
Tier 4 (Max)	20,000	50,000	40M
Enterprise	Custom	Custom	Custom

SDKs — Quick Reference

Language	Package	Import
Python	pip install anthropic	import anthropic
TypeScript	npm install @anthropic-ai/sdk	import Anthropic from '@anthropic-ai/sdk'
Java	Maven/Gradle: com.anthropic:anthropic-java	import com.anthropic.*
Go	go get github.com/anthropics/anthropic-sdk-go	import "github.com/anthropics/anthropic-sdk-go"

Where Next

• Claude Code — the full agentic coding tool
• Workflows & Best Practices — prompt engineering and optimization
• Enterprise & Deployment — AWS, GCP, Teams, compliance

For cross-model comparison (GPT, Gemini, DeepSeek), see the Models Decision Guide.