The LangSmith SDKs are the Python and JavaScript/TypeScript client libraries for the LangSmith platform, which is used to debug, evaluate, and monitor language-model applications and agents. LangSmith works with any LLM application and has a native integration with the LangChain Python and JavaScript libraries.
Tracing starts by setting LANGSMITHTRACING=true and an API key (plus a workspace ID for organization-scoped keys) and then either wrapping a client, for example wrapopenai around an OpenAI client, or decorating functions with traceable. Calls made through the wrapped client are recorded as runs on the platform. The same pattern is available in JavaScript through wrapOpenAI and traceable.
Features
- Two languages: a Python package (langsmith) and a JavaScript/TypeScript package with matching APIs
- Client wrappers: wrap_openai / wrapOpenAI capture OpenAI SDK calls without changing application code
- Function tracing: the traceable decorator records arbitrary functions as runs
- Environment configuration: tracing toggled and authenticated through LANGSMITHTRACING, LANGSMITHAPIKEY, and LANGSMITHWORKSPACE_ID
- LangChain integration: native tracing for LangChain Python and JS applications
- Evaluation and monitoring: the SDK is the client for the platform's evaluation and monitoring features, documented in the LangSmith docs and cookbook
Integrates with
Integrated by
Deep Agents
Opinionated, model-agnostic agent harness on LangGraph with sub-agents, filesystem, and context management built in
DeerFlow
DeerFlow 2.0 is ByteDance's open-source super-agent harness.
Dexter
Autonomous financial research agent that thinks, plans, and learns as it works. It performs analysis using task planning, self-reflection, and real-time market data.
GPT Researcher
Deep-research agent that plans questions, crawls 20+ sources in parallel, and compiles cited reports
