Build applications on top of language models.
LangChain gives you a small set of composable pieces — prompts, models, retrievers, tools — and one interface that ties them together.
chain = prompt | model | StrOutputParser()
chain.invoke({"topic": "vector search"})Compose, don't glue
Prompts, models, retrievers and parsers share one interface. Pipe them together and the result behaves like every other piece.
Ground every answer
Bring your own documents, split them, embed them and hand the model only the context the question needs.
Ship with confidence
Streaming, tracing and evaluation are part of the framework, not an afterthought you bolt on before launch.
Start with the docs
All 8 pagesGet started
Installation
Install LangChain and wire up your first model provider.
Get started
Quickstart
Build a prompt, model and parser pipeline in under ten lines.
Core concepts
Core concepts
Runnables, messages, prompts and the composition model.
Build
Retrieval augmented generation
Ground model answers in your own documents with a vector store.
From the blog
All postsRetrieval · 18 September 2026
Retrieval patterns that hold up in production
Naive similarity search gets you a demo. These four adjustments are what carried our retrieval pipeline through real traffic.
Agents · 27 August 2026
Designing tools agents can actually use
Most agent failures are not reasoning failures. They are interface failures, and the fix is in your function signatures.
Engineering · 14 July 2026
Streaming UX that actually feels fast
Time to first token is the metric your users feel. Here is how to spend it well across retrieval, tools and generation.