What it does
LangChain is an MIT-licensed framework for building agents and applications powered by large language models. It gives models, embeddings, vector stores and other components a standard interface, so parts can be chained together and swapped as the underlying technology changes. The repository describes working at whatever level suits you, from high-level chains for a quick start down to low-level components for fine control. Integrations cover model providers, tools, vector stores and retrievers. The quickstart installs the package with uv and initialises a chat model with a single call. The page places LangChain inside a wider ecosystem. Deep Agents covers agents with planning, subagents and file system use. LangGraph is a lower-level framework for controllable agent workflows. LangSmith provides evaluations, observability and debugging, and LangSmith Deployment runs long-lived, stateful agents. A separate library, LangChain.js, covers JavaScript and TypeScript.
Capabilities
- Standard interface for chat models, embeddings and vector stores
- Large library of integrations with model providers, tools and retrievers
- Swap models without rebuilding the application
- Modular components, from high-level chains to low-level control
- Monitoring, evaluation and debugging through integrations such as LangSmith
When to use it
- Building an LLM-powered app that may need to change model providers later
- Prototyping agents or retrieval workflows from modular components
- Starting simple, then moving to LangGraph when you need more controllable agent orchestration
