Capability · Framework — observability

LangSmith

LangSmith is the production platform for LangChain/LangGraph applications (and any OpenTelemetry-compatible LLM app). It provides distributed tracing of chains, agents, and tool calls; dataset collection from production traces; online and offline evaluations with LLM-as-judge; prompt versioning and playground; and dashboards for cost, latency, and quality. Self-hosted and cloud options are available.

Framework facts

Category: observability
Language: Python / TypeScript
License: commercial
Repository: https://github.com/langchain-ai/langsmith-sdk

Install

pip install langsmith
# or
npm install langsmith

Quickstart

import os
from langsmith import traceable

os.environ['LANGSMITH_TRACING'] = 'true'
os.environ['LANGSMITH_API_KEY'] = '...'

@traceable
def answer(q: str) -> str:
    return f'echo: {q}'

answer('hi')  # appears in LangSmith UI

Alternatives

Langfuse — open-source alternative
Arize Phoenix — OTel-based open source
Helicone — proxy-based observability
Braintrust — evals + monitoring

Frequently asked questions

Do I need to use LangChain to use LangSmith?

No. LangSmith accepts OpenTelemetry traces and has direct SDK support. You can instrument any LLM app — OpenAI SDK, Anthropic SDK, vanilla Python — and get the same tracing, datasets, and evals UX.

What does LangSmith cost?

LangSmith has a free Developer plan with a monthly trace cap, paid Plus/Enterprise tiers for teams, and self-hosted enterprise licensing for on-prem or VPC deployments. Check the pricing page for current limits.

Sources

LangSmith — docs — accessed 2026-04-20
LangSmith — site — accessed 2026-04-20