How to Intermediate · 3 min read

How to use LangSmith with LangGraph

Q: How to use LangSmith with LangGraph

Use the langsmith Python SDK to create a Client and decorate your LangGraph node functions with @client.traceable to automatically trace state transitions. Compile your StateGraph into an app and invoke it normally; LangSmith captures detailed traces for observability.

Quick answer

Use the langsmith Python SDK to create a Client and decorate your LangGraph node functions with @client.traceable to automatically trace state transitions. Compile your StateGraph into an app and invoke it normally; LangSmith captures detailed traces for observability.

PREREQUISITES

Python 3.8+
pip install langsmith langgraph
LangSmith API key set in environment variable LANGSMITH_API_KEY

Setup

Install the required packages and set your LangSmith API key in the environment.

Install packages: pip install langsmith langgraph
Set environment variable: export LANGSMITH_API_KEY=your_api_key (Linux/macOS) or set LANGSMITH_API_KEY=your_api_key (Windows)

bash

pip install langsmith langgraph

Step by step

This example shows how to create a LangGraph stateful agent and use LangSmith to trace its execution by decorating node functions with @client.traceable.

python

import os
from langsmith import Client, traceable
from langgraph.graph import StateGraph, END
from typing import TypedDict

# Initialize LangSmith client
client = Client(api_key=os.environ["LANGSMITH_API_KEY"])

# Define the state type
class State(TypedDict):
    messages: list[str]

# Define a traced node function
@client.traceable
def add_greeting(state: State) -> State:
    new_messages = state["messages"] + ["Hello from LangGraph with LangSmith!"]
    return {"messages": new_messages}

# Create the graph
graph = StateGraph(State)
graph.add_node("greet", add_greeting)
graph.set_entry_point("greet")
graph.add_edge("greet", END)

# Compile the graph into an app
app = graph.compile()

# Invoke the app
result = app.invoke({"messages": []})
print("Final state messages:", result["messages"])

output

Final state messages: ['Hello from LangGraph with LangSmith!']

Common variations

You can trace multiple nodes by decorating each with @client.traceable. For asynchronous LangGraph nodes, use @client.traceable on async functions. LangSmith automatically captures traces for all decorated nodes, enabling detailed observability.

python

import asyncio

@client.traceable
async def async_node(state: State) -> State:
    await asyncio.sleep(0.1)
    new_messages = state["messages"] + ["Async node executed"]
    return {"messages": new_messages}

# Add async node example
graph.add_node("async_node", async_node)
graph.add_edge("greet", "async_node")
graph.add_edge("async_node", END)

app = graph.compile()

# Run async invocation
async def main():
    result = await app.invoke({"messages": []})
    print("Async final messages:", result["messages"])

asyncio.run(main())

output

Async final messages: ['Hello from LangGraph with LangSmith!', 'Async node executed']

Troubleshooting

If you see Missing LANGSMITH_API_KEY errors, ensure your environment variable is set correctly before running your script.
If traces do not appear in the LangSmith dashboard, verify network connectivity and that your API key is valid.
For issues with async nodes, confirm you are using Python 3.8+ and that your functions are properly declared async.

✅

Key Takeaways

Use langsmith.Client and @client.traceable to instrument LangGraph nodes for tracing.
Compile your StateGraph into an app and invoke it normally; LangSmith captures detailed execution traces.
Support for async LangGraph nodes is seamless with @client.traceable on async functions.
Ensure your LANGSMITH_API_KEY environment variable is set to enable tracing.
LangSmith enhances debugging and observability of complex LangGraph stateful agents.

Verified 2026-04

Verify ↗