What's the difference between the Claude Agent SDK and LangChain?

The Claude Agent SDK is, in Anthropic’s own words, “a library that runs the Claude Code binary,” so it only runs Claude; LangChain is a model-agnostic orchestration library that reaches Claude, GPT, Gemini, or any other supported provider through one unified interface.

That difference shows up in what each one ships with. The Claude Agent SDK inherits Claude Code’s own capabilities wholesale: built-in tools for reading, writing, and editing files, running commands, and searching the web, plus hooks, subagents, MCP support, permissions, and sessions you can resume or fork. None of that is optional or swapped out for a different model. LangChain instead gives you “a single, unified API to work with models from any provider,” over a thousand integrations, and builds its own agents with create_agent, its current agent-construction primitive, formerly named create_react_agent.

That old name is a real clue: LangChain’s agent builder runs the ReAct loop underneath, a reasoning step interleaved with each tool call. The Claude Agent SDK’s loop is Claude Code’s own and isn’t exposed as a separate, swappable reasoning primitive at all.