LangSmith

LangSmith is LangChain's platform for tracing, monitoring and evaluating AI agents and LLM apps, with SDKs for Python, TypeScript, Go and Java.

LangSmith is an observability and evaluation platform for teams building agents and LLM applications, whether or not they use LangChain. Tracing records each step of a run, including model calls, tool calls and multi-turn threads, so developers can find what hurt latency, cost or answer quality. Monitoring dashboards track token usage, latency, error rates and cost, score production traffic with online LLM-as-judge and code evaluations, and send alerts through webhooks or PagerDuty. Insights clusters traces automatically to surface usage patterns and common failure modes. It works with the OpenAI and Anthropic SDKs, the Vercel AI SDK, LlamaIndex and OpenTelemetry pipelines. The Plus plan adds agent deployment and unlimited seats, and Enterprise customers can self-host it in their own cloud.

LangSmith at a glance

Best for: Software engineer
Runs on: Web, API, self-hosted
Pricing: Developer free with 5,000 traces/month and 1 seat; Plus $39/seat/month with 10,000 traces; Enterprise custom; extra traces pay as you go
Released: July 2023
Category: AI coding tool
Website: langchain.com

Key features

Step-by-step traces of model and tool calls
Dashboards for cost, latency, token usage and error rates
Online LLM-as-judge and code evaluations
Insights that cluster traces into usage patterns and failure modes
Webhook and PagerDuty alerts
OpenTelemetry support and SDKs for Python, TypeScript, Go and Java
Self-hosted and hybrid deployment on the Enterprise plan

Use it for

Find the step that makes an agent slow or expensive
Score production answers for quality with automatic evals
Test a prompt or model change before release
Alert the team when error rates cross a threshold

Companions

Tools that articles about LangSmith recommend using with it, each with its job.

LangChain — Built to work with LangChain.
LangGraph — Durable execution, checkpoints and human approval under create_agent.
Pinecone — Managed retrieval for the fastest path to production.
Qdrant — Self-hosted vectors when cost or data residency matter.
Ollama — Fully self-hosted inference paired with Qdrant.

Alternatives to LangSmith

LangWatch — Runs simulations and continuous evaluations to make AI agents reliable.
LangChain4j — Integrates LLMs and vector stores directly into Java applications.
Langfuse — It connects tracing, prompt management, and evaluations in one open platform.
Langfa.st — Compare model responses side-by-side to validate prompts before deployment.
Langroid — Simplifies multi-agent orchestration and app development with intuitive abstractions.
Agent Smith — It reduces token overspending by routing routine work to lighter models.
Tracely — Turns production failures into regression tests that block pull requests.
Loushang — It provides persistent sessions, model routing, and method-guided code delivery.
OpenLIT — It enables comprehensive monitoring, evaluation, and management of AI agents via OpenTelemetry.
Lunary — It enables teams to monitor LLM performance, debug agents, and manage prompts.
OpenCode — Works with 75+ model providers, local models, and existing subscriptions
LiteLLM — One OpenAI-format API for 100+ LLM providers

All LangSmith alternatives