LangChainLangChain is a Python framework for building agents and LLM applications by chaining interoperable components and third-party integrations. It provides a standard interface for models, embeddings, vector stores, retrievers, and tools, so you can swap a model or a data source without rebuilding the application. Its agent entry point is a minimal, configurable agent harness that you compose from a model, tools, a prompt, and middleware.
The framework works at several levels of abstraction, from high-level chains for a quick start to low-level components for fine-grained control. Its integrations library connects LLMs to chat and embedding model providers, tools and toolkits, vector stores, and retrievers.
LangChain can be used on its own or with the rest of the LangChain ecosystem. Deep Agents is a higher-level package built on LangChain for agents that plan, use subagents, and work with a file system. LangGraph is the lower-level orchestration framework for controllable agent workflows. LangSmith covers evals, observability, debugging, and deployment for long-running, stateful agents. LangChainJS is the equivalent JavaScript and TypeScript library.
Features
- Standard model interface: one interface across model providers, so models can be swapped as requirements change
- Agent harness: compose an agent from a model, tools, a prompt, and middleware
- Integrations library: chat and embedding models, tools and toolkits, vector stores, and retrievers
- Layered abstractions: high-level chains for quick starts and low-level components for fine-grained control
- Deep Agents: a higher-level package for planning, subagents, and file-system use
- Ecosystem fit: works with LangGraph for orchestration and LangSmith for evals, observability, and deployment
