LangChain
The original Python/TypeScript framework for building LLM applications — provides abstractions for chains, agents, tool use, memory, and retrieval. In 2026, mostly superseded by LangGraph for new projects.
LangChain was the dominant agent framework of 2023–2024. It provides building blocks — chains, agents, tools, memory, retrievers — for composing LLM applications without rebuilding the primitives.
In 2026, the LangChain team has largely shifted focus to [LangGraph](/glossary/langgraph) for new agent work. LangChain proper is still maintained but feels increasingly heavy for what modern vendor SDKs (Anthropic, OpenAI) provide directly. Most new production agents skip LangChain in favor of LangGraph or vendor-native SDKs.
Where LangChain still earns its keep: research and prototyping where its breadth of integrations (200+ tools, vector stores, LLM providers) saves time. For production, the abstractions often add complexity without much benefit over direct SDK use.
Frequently asked
Should I use LangChain in 2026?+
Probably not for new production work. Use [LangGraph](/glossary/langgraph) (same team, better foundation) or a vendor SDK (Anthropic Agent SDK, OpenAI Agents SDK). LangChain is fine for prototyping but rarely the right production choice.
What replaced LangChain?+
LangGraph for stateful agents. CrewAI for role-based multi-agent. AutoGen for research-style conversational multi-agent. Anthropic Agent SDK and OpenAI Agents SDK for vendor-native development.