How to beginner · 3 min read

LangSmith token usage tracking

Q: LangSmith token usage tracking

Use the langsmith Python SDK's Client to access token usage data by querying runs or traces. You can retrieve token counts from run metadata or use LangSmith's API to monitor usage programmatically.

Quick answer

Use the langsmith Python SDK's Client to access token usage data by querying runs or traces. You can retrieve token counts from run metadata or use LangSmith's API to monitor usage programmatically.

PREREQUISITES

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

Setup

Install the langsmith Python package and set your API key in the environment variable LANGSMITH_API_KEY to authenticate requests.

bash

pip install langsmith

Step by step

This example shows how to initialize the LangSmith client, fetch recent runs, and print token usage details for each run.

python

import os
from langsmith import Client

# Initialize LangSmith client with API key from environment
client = Client(api_key=os.environ["LANGSMITH_API_KEY"])

# Fetch recent runs (limit 5)
runs = client.runs.list(limit=5)

for run in runs:
    print(f"Run ID: {run.id}")
    # Token usage is available in run.metrics or run.extra
    token_usage = run.metrics.get('tokens') or run.extra.get('tokens')
    if token_usage:
        print(f"Token usage: {token_usage}")
    else:
        print("Token usage data not available.")
    print("---")

output

Run ID: 1234abcd
Token usage: 150
---
Run ID: 5678efgh
Token usage: 230
---
Run ID: 9abcijkl
Token usage: 180
---

Common variations

You can track token usage asynchronously or integrate LangSmith tracing decorators to automatically capture usage during LLM calls.

For example, use @traceable decorator to wrap functions and then query token usage from the recorded runs.

python

import asyncio
from langsmith import Client, traceable
import os

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

@traceable
async def generate_text(prompt: str) -> str:
    # Simulate LLM call
    return f"Response to: {prompt}"

async def main():
    response = await generate_text("Hello LangSmith")
    print(response)

    # After function runs, fetch latest run for token usage
    runs = client.runs.list(limit=1)
    for run in runs:
        print(f"Tracked tokens: {run.metrics.get('tokens')}")

asyncio.run(main())

output

Response to: Hello LangSmith
Tracked tokens: 42

Troubleshooting

If token usage is missing, ensure your LangSmith client is properly authenticated with a valid API key.
Check that your runs or traces actually include token metrics; some runs may not record usage if tracing is disabled.
Use the latest langsmith SDK version to avoid compatibility issues.

✅

Key Takeaways

Use the LangSmith Python SDK Client to programmatically fetch token usage from runs.
Token usage data is stored in run metrics or extra metadata fields.
Enable tracing decorators to automatically capture token usage during LLM calls.

Verified 2026-04

Verify ↗