Evals & observability
LangSmith
Tracing, datasets, and evaluation from the LangChain team; works with or without LangChain.
- Category
- Evals & observability
- Pricing
- Free tier
- Runs
- Hosted
- Languages
- Python, TypeScript
- Interface
- SDK, Dashboard
- Source
- Proprietary
What LangSmith is
LangSmith provides tracing, dataset management, and evaluation, and while it comes from the LangChain team it does not require LangChain. If you do use LangChain or LangGraph, instrumentation is close to automatic and nested chains and agent steps come through as a readable tree.
Best for
LangChain and LangGraph users, who get the deepest integration for the least work.
Consider something else if
Hosted only, so traces containing user data leave your infrastructure.
LangSmith alternatives
The closest options in evals & observability, on the axes that actually separate them.
| Tool | Best for | Pricing | Runs |
|---|---|---|---|
| LangSmith | LangChain and LangGraph users, who get the deepest integration for the least work. | Free tier | Hosted |
| Langfuseopen source | Teams who need traces containing customer data to stay inside their own infrastructure. | Open source | Hosted or self-hosted |
| Braintrust | Teams treating prompt changes as changes that need tests, with evals wired into CI. | Free tier | Hosted |
| Phoenixopen source | Teams standardised on OpenTelemetry, and anyone debugging retrieval quality interactively. | Open source | Hosted or self-hosted |
| Heliconeopen source | Getting cost and latency visibility across a codebase quickly, especially one you do not want to instrument. | Free tier | Hosted or self-hosted |
| W&B Weaveopen source | Organisations already on Weights & Biases who want one place for all model work. | Free tier | Hosted |
Choosing within evals & observability
Tracing or evaluation first
Some tools lead with tracing and add evals; others are eval-first with tracing attached. If you are debugging today, start with tracing. If you are about to change a prompt that is already in production, start with evals.
How it gets your data
A proxy needs a one-line base-URL change and captures everything. An SDK needs instrumentation and captures more structure. OpenTelemetry-based tools slot into tracing you may already run.
Questions
What is LangSmith?
LangSmith provides tracing, dataset management, and evaluation, and while it comes from the LangChain team it does not require LangChain. If you do use LangChain or LangGraph, instrumentation is close to automatic and nested chains and agent steps come through as a readable tree. It is a commercial product and hosted.
What are the alternatives to LangSmith?
The closest alternatives are Langfuse, Braintrust, Phoenix, Helicone, W&B Weave. They sit in the same category — evals & observability — and differ mainly on hosting model, pricing shape, and how much they abstract away.
Is LangSmith the right choice?
LangChain and LangGraph users, who get the deepest integration for the least work. The main caveat: Hosted only, so traces containing user data leave your infrastructure.
Whatever you build on, the model is the line item that scales. See what each one costs per million tokens, or price your own workload.