Claude SDK

January 30, 2026

The Claude SDK is Anthropic's official client library for building applications with Claude models. Available in Python and TypeScript, it provides direct access to Claude's capabilities including tool use, streaming, and agent workflows.

Core Features

Simple API Access

from anthropic import Anthropic

client = Anthropic()
message = client.messages.create(
    model="claude-sonnet-4-5-20250929",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Hello, Claude"}
    ]
)
print(message.content[0].text)

Streaming Responses

with client.messages.stream(
    model="claude-sonnet-4-5-20250929",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Write a story"}]
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

Tool Use

Claude can call functions you define:

tools = [
    {
        "name": "get_weather",
        "description": "Get the current weather in a location",
        "input_schema": {
            "type": "object",
            "properties": {
                "location": {"type": "string"}
            },
            "required": ["location"]
        }
    }
]

response = client.messages.create(
    model="claude-sonnet-4-5-20250929",
    max_tokens=1024,
    tools=tools,
    messages=[{"role": "user", "content": "What's the weather in Tokyo?"}]
)

Claude Agent SDK

Beyond the base API, Anthropic offers agent-focused tooling. As noted on Twitter:

We have open-sourced our Claude Agents SDK Container repo which allows you to deploy the new Claude Agents SDK in a container that uses your Anthropic Max plan. Serve agents from anywhere.

Container Deployment

The Claude Agent SDK can be packaged into containers for deployment anywhere:

FROM python:3.11-slim

COPY requirements.txt .
RUN pip install -r requirements.txt

COPY agent.py .
CMD ["python", "agent.py"]

This enables:

Running agents on your infrastructure
Integration with event triggers
Connection to MCP servers
Scaling based on demand

Integration with MCP

Claude SDK integrates natively with MCP servers:

from anthropic import Anthropic

client = Anthropic()

# Connect to MCP servers for extended capabilities
response = client.messages.create(
    model="claude-sonnet-4-5-20250929",
    max_tokens=1024,
    mcp_servers=["file://./my-mcp-server"],
    messages=[{"role": "user", "content": "Search my files for..."}]
)

Prompt Caching

Reduce costs and latency by caching common prompt prefixes:

response = client.messages.create(
    model="claude-sonnet-4-5-20250929",
    max_tokens=1024,
    system=[
        {
            "type": "text",
            "text": "You are a helpful assistant...",
            "cache_control": {"type": "ephemeral"}
        }
    ],
    messages=[{"role": "user", "content": "..."}]
)

Computer Use (Beta)

Claude can interact with computer interfaces:

tools = [
    {
        "type": "computer_20241022",
        "name": "computer",
        "display_width_px": 1024,
        "display_height_px": 768,
    }
]

response = client.messages.create(
    model="claude-sonnet-4-5-20250929",
    max_tokens=1024,
    tools=tools,
    messages=[{"role": "user", "content": "Open the browser and go to..."}]
)

TypeScript SDK

Same capabilities in TypeScript:

import Anthropic from '@anthropic-ai/sdk';

const anthropic = new Anthropic();

const message = await anthropic.messages.create({
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [
    { role: "user", content: "Hello, Claude" }
  ],
});

When to Use the Claude SDK

Good fit:

Building Claude-specific applications
Need direct API access without abstractions
Using Claude-specific features (prompt caching, computer use)
Deploying containerized agents

Consider alternatives when:

Need multi-provider flexibility (use Vercel AI SDK)
Building complex orchestration (use LangGraph)
Want higher-level abstractions

Claude Code Integration

The SDK powers Claude Code, Anthropic's coding assistant. Features like:

File system access
Tool execution
Multi-turn conversations
Context management

Are all built on the Claude SDK primitives.

Sources

Anthropic API Documentation — Official docs
GitHub: anthropics/anthropic-sdk-python — Python SDK
GitHub: anthropics/anthropic-sdk-typescript — TypeScript SDK