What is the LangSmith platform?

LangSmith is LangChain's hosted observability and evaluation platform for LLM applications. It automatically captures traces — the full execution tree of every chain, agent step, LLM call, retriever hit, and tool invocation — so you can inspect exactly what happened during a run, including prompts sent, completions received, latency at each step, and token usage.

Enabling LangSmith requires just two environment variables:

export LANGCHAIN_TRACING_V2=true
export LANGCHAIN_API_KEY=ls__your_key_here
export LANGCHAIN_PROJECT=my-project   # optional, groups traces

No code changes are required — every LangChain component automatically sends traces once these are set. LangSmith's main capabilities include:

• Tracing — visualise the full execution tree of any run
• Datasets & Evaluations — build golden datasets and run LLM-as-judge or custom evaluators to benchmark prompt changes
• Playground — edit prompts inline and replay traces to test changes
• Monitoring — dashboards for latency, error rates, and cost over time in production
• Annotation queues — route interesting traces to human reviewers for labelling and feedback