Trace with Anthropic

• From the Claude Platform: The Claude Platform sends traces directly from Anthropic’s servers. You configure it in the Claude Console, with no changes to your agent code.
• From your agent: Your agent sends the traces, in one of two ways:
  • Claude SDKs: The Claude SDKs emit OpenTelemetry (OTel) spans, and an OTel exporter library sends them to LangSmith.
  • LangSmith SDK wrapper: You wrap the Anthropic client with the LangSmith SDK (Python or TypeScript). You can combine the traces with @traceable spans for the rest of your code in a single trace.

​

Trace from the Claude Platform

• The admin role, or an equivalent permission, in your Claude Console organization. Without it, you cannot view or manage destinations.
• A LangSmith API key. You can use a personal access token, or a service key when you want tracing to be independent of your user account. To create a key, see Create an account and API key.
• In an agent-based workspace, the selected agent environment’s tracer_session_name, which goes in the destination’s Project field.
• For self-hosted LangSmith, an endpoint that the Claude Platform can reach: HTTPS on port 443, at a hostname that resolves to a public IP address.

​

Step 1. Add LangSmith as a destination

1. In the Claude Console, open the OpenTelemetry settings page of your organization and click Add destination.
2. In the Provider drop-down, select LangSmith. The form changes to show the LangSmith fields.
3. Complete the fields:
   • Name: (Optional) A label for the destination, such as LangSmith.
   • Region: The region of your LangSmith deployment: US, EU, APAC, or US on AWS. For self-hosted LangSmith, select Self-hosted. In LangSmith URL, enter your LangSmith API URL (LANGSMITH_ENDPOINT) plus /otel. For the URL on each deployment type, see Find your OTLP endpoint.
   • API key: Your LangSmith API key.
   • Project: The environment’s tracer_session_name to send traces to. Set it explicitly in an agent-based workspace. If omitted, the integration uses default.
   • LangSmith workspace ID: (Optional) The ID of the LangSmith workspace to send traces to. If you leave it empty, traces go to the workspace that the API key belongs to. Organization-scoped service keys require it. To find the ID, see Create an account and API key.
   • Content mode: Select Full content to include prompts and responses in the spans. With Metadata only, spans omit the input and output content.
   • Workspaces: The Claude workspaces that send traces to this destination: All, Selected, or All except. For details on each option, see Choose workspaces for a destination in the Anthropic docs.
4. Click Add destination.

• x-api-key: The API key.
• Langsmith-Project: The Project, if you set one.
• X-Tenant-Id: The LangSmith workspace ID, if you set one.

​

Step 2. Check that traces arrive

1. Send a Messages API request or run a Claude Managed Agents session in a Claude workspace that uses the destination.
2. On the Claude Console OpenTelemetry settings page, check that the Delivery column of the destination shows Healthy. This can take about a minute.
3. In LangSmith, select the agent and environment whose tracer_session_name you entered in the Project field.

​

Troubleshoot delivery

​

Trace from your agent

​

Use the Claude SDKs

1. Install or upgrade the Claude SDK and register a tracer provider.
2. Configure an OTel exporter with your LangSmith API URL (LANGSMITH_ENDPOINT) plus /otel as the base endpoint. For the URL on each deployment type, see Find your OTLP endpoint. Set the following request headers:
   • x-api-key: Your LangSmith API key.
   • Langsmith-Agent-Id: The registered agent identifier.
   • Langsmith-Agent-Environment: The environment, such as local.
   • X-Tenant-Id: (Optional) The ID of the LangSmith workspace to send traces to. Without this header, traces go to the workspace that the API key belongs to. Organization-scoped service keys require it.
   For example, with the standard OTel environment variables:
   export OTEL_EXPORTER_OTLP_ENDPOINT=https://api.smith.langchain.com/otel
   export OTEL_EXPORTER_OTLP_HEADERS="x-api-key=<your-langsmith-api-key>,Langsmith-Agent-Id=<your-agent-id>,Langsmith-Agent-Environment=local"
   
   For more information on exporter setup, see Trace a non-LangChain agent.
3. (Optional) By default, the Claude SDKs record metadata only. To include prompts and responses in the spans, set the content mode. See Configure the SDK in the Anthropic docs.

​

Use the LangSmith SDK wrapper

import anthropic
from langsmith import traceable
from langsmith.wrappers import wrap_anthropic

client = wrap_anthropic(anthropic.Anthropic())

@traceable(run_type="tool", name="Retrieve Context")
def my_tool(question: str) -> str:
  return "During this morning's meeting, we solved all world conflict."

@traceable(name="Chat Pipeline")
def chat_pipeline(question: str):
  context = my_tool(question)
  messages = [
      { "role": "user", "content": f"Question: {question}\nContext: {context}"}
  ]
  message = client.messages.create(
      model="claude-sonnet-5-5",
      messages=messages,
      max_tokens=1024,
      system="You are a helpful assistant. Please respond to the user's request only based on the given context."
  )
  return message

chat_pipeline("Can you summarize this morning's meetings?")

import Anthropic from "@anthropic-ai/sdk";
import { traceable } from "langsmith/traceable";
import { wrapAnthropic } from "langsmith/wrappers/anthropic";

const client = wrapAnthropic(new Anthropic());

const myTool = traceable(async (question: string) => {
  return "During this morning's meeting, we solved all world conflict.";
}, { name: "Retrieve Context", run_type: "tool" });

const chatPipeline = traceable(async (question: string) => {
  const context = await myTool(question);
  const messages = [
      { role: "user", content: `Question: ${question}\nContext: ${context}` }
  ];
  const message = await client.messages.create({
      model: "claude-sonnet-5-5",
      messages: messages,
      max_tokens: 1024,
      system: "You are a helpful assistant. Please respond to the user's request only based on the given context."
  });
  return message;
}, { name: "Chat Pipeline" });

await chatPipeline("Can you summarize this morning's meetings?");