Trace Claude Agent SDK applications

The Claude Agent SDK is an SDK for building agentic applications with Claude. LangSmith provides native integration with the Claude Agent SDK to automatically trace your agent executions, tool calls, and interactions with Claude models.

Installation

Install the LangSmith integration for Claude Agent SDK

uv

uv add "langsmith[claude-agent-sdk]"

pip install langsmith[claude-agent-sdk]

pnpm add @anthropic-ai/claude-agent-sdk langsmith zod

npm install @anthropic-ai/claude-agent-sdk langsmith zod

Setup

Set your API keys:

shell

export LANGSMITH_TRACING=true
export LANGSMITH_ENDPOINT=https://api.smith.langchain.com
export LANGSMITH_API_KEY=<your_langsmith_api_key>
export LANGSMITH_PROJECT=<your_langsmith_project>

export ANTHROPIC_API_KEY=<your_anthropic_api_key>

LANGSMITH_TRACING=true
LANGSMITH_ENDPOINT=https://api.smith.langchain.com
LANGSMITH_API_KEY=<your_langsmith_api_key>
LANGSMITH_PROJECT=<your_langsmith_project>

ANTHROPIC_API_KEY=<your_anthropic_api_key>

You can find your LangSmith API key and project name in the LangSmith UI under Settings.

For an Anthropic API key, refer to the Claude console.

Quickstart

To enable LangSmith tracing for your Claude Agent SDK application, call configure_claude_agent_sdk() at the start of your application:

Python

import asyncio
from typing import Any

from claude_agent_sdk import (
    ClaudeAgentOptions,
    ClaudeSDKClient,
    create_sdk_mcp_server,
    tool,
)
from langsmith.integrations.claude_agent_sdk import configure_claude_agent_sdk

configure_claude_agent_sdk()


@tool(
    "get_weather",
    "Gets the current weather for a given city",
    {"city": str},
)
async def get_weather(args: dict[str, Any]) -> dict[str, Any]:
    city = args["city"]
    weather_data = {
        "San Francisco": "Foggy, 62°F",
        "New York": "Sunny, 75°F",
        "London": "Rainy, 55°F",
        "Tokyo": "Clear, 68°F",
    }
    weather = weather_data.get(city, "Weather data not available")
    return {"content": [{"type": "text", "text": f"Weather in {city}: {weather}"}]}


async def main() -> None:
    weather_server = create_sdk_mcp_server(
        name="weather",
        version="1.0.0",
        tools=[get_weather],
    )

    options = ClaudeAgentOptions(
        model="claude-sonnet-4-5-20250929",
        system_prompt="You are a friendly travel assistant who helps with weather information.",
        mcp_servers={"weather": weather_server},
        allowed_tools=["mcp__weather__get_weather"],
    )

    async with ClaudeSDKClient(options=options) as client:
        await client.query("What's the weather like in San Francisco and Tokyo?")

        async for message in client.receive_response():
            print(message)


if __name__ == "__main__":
    asyncio.run(main())

import * as originalSdk from '@anthropic-ai/claude-agent-sdk';

import { wrapClaudeAgentSDK } from 'langsmith/experimental/anthropic';
import { z } from 'zod/v4';

const sdk = wrapClaudeAgentSDK(originalSdk);

const getWeather = sdk.tool(
  'get_weather',
  'Gets the current weather for a given city',
  {
    city: z.string(),
  },
  async ({ city }) => {
    const weatherData: Record<string, string> = {
      'San Francisco': 'Foggy, 62°F',
      'New York': 'Sunny, 75°F',
      London: 'Rainy, 55°F',
      Tokyo: 'Clear, 68°F',
    };
    const weather = weatherData[city] ?? 'Weather data not available';
    return {
      content: [{ type: 'text' as const, text: weather }],
    };
  }
);

const weatherServer = sdk.createSdkMcpServer({
  name: 'weather',
  version: '1.0.0',
  tools: [getWeather],
});

const query = sdk.query({
  prompt: "What's the weather like in San Francisco and Tokyo?",
  options: {
    model: 'claude-sonnet-4-5-20250929',
    systemPrompt:
      'You are a friendly travel assistant who helps with weather information.',
    mcpServers: { weather: weatherServer },
    allowedTools: ['mcp__weather__get_weather'],
  },
});

for await (const chunk of query) {
  console.log(chunk);
}

Once configured, all Claude Agent SDK operations will be automatically traced to LangSmith, including:

• Agent queries and responses
• Tool invocations and results
• Claude model interactions
• MCP server operations

---

Connect these docs to Claude, VSCode, and more via MCP for real-time answers.

Edit this page on GitHub or file an issue.