LangChain logo
Serving and AI platform

LangChain

Framework for building LLM applications and agents. LangChain is an open-source framework, in Python and JavaScript, for building applications on language models. It provides common interfaces for models, retrievers, vector stores and tools, and LangGraph, from the same team, adds stateful agent workflows. LangSmith is the company's commercial tracing and evaluation product.

When it is the right choice

  • You want to switch between model providers behind one interface
  • You are building RAG or agent workflows and want ready-made integrations
  • You need stateful, multi-step agents with human approval steps (LangGraph)
  • Not always needed: for a few direct model calls, the provider SDK is easier to read and debug

How we use it

  • Use LangChain or LangGraph where they reduce code, and plain SDK calls where they do not
  • Instrument chains and agents with OpenTelemetry so every step is traceable
  • Build evaluation datasets and run them in CI before a prompt or model change ships
  • Deploy the application as a container with the same pipeline as your other services
Works well with

Often used together

Questions

LangChain: frequently asked

Is LangChain production ready?

Many teams run it in production. The key is pinning versions, tracing every step and testing changes, because it is a large and fast-moving library.

Do we need LangSmith?

No. LangSmith is helpful for tracing and evaluation, but OpenTelemetry-based tooling can cover the same ground if you prefer to keep data in your own stack.

Running LangChain, or deciding whether to? Tell us what you have; the first consultation is free.

Let's talk »