LangChain

LangChain is an open-source Python and JavaScript framework that has become the dominant development toolkit for building applications powered by large language models, providing composable abstractions for prompts, chains, agents, memory, and tool integration that reduce the code required to build complex LLM-powered applications from hundreds of lines to dozens. Its chain abstraction sequences multiple LLM calls and processing steps into reusable pipelines: a retrieval-augmented generation (RAG) chain retrieves relevant documents from a vector database, formats them into a prompt, calls an LLM, and returns a grounded response in a few lines of configuration. The current create_agent API and harness abstraction enable LLMs to use external tools, web search, calculators, code execution, APIs, and decide which tools to call based on the task, implementing the ReAct and tool-use patterns without manual orchestration code, while the separate LangGraph framework adds stateful, graph-based orchestration for more complex multi-agent workflows. LangChain integrates with 100+ LLM providers, 50+ vector databases, and hundreds of external tool APIs out of the box.

The open-source framework has a $0 license fee; model usage and hosting costs remain separate. LangSmith, a separately priced companion observability and deployment platform, traces every LLM call, measures latency and token usage, and enables debugging of complex multi-step chains. The framework's breadth has made it the starting point for most LLM application development projects, though teams with high-scale production needs often migrate to lighter-weight custom implementations as they optimize. Best suited for teams building LLM applications and RAG systems who want community support, extensive integrations, and production-tested patterns rather than building from primitives.