What's the difference between MCP and LangChain?
MCP is an open protocol for exposing tools and data sources to any compatible client, not tied to one framework; LangChain is an orchestration library that, since version 1.4, ships a built-in client for consuming an MCP server’s tools rather than an alternative to MCP itself.
MCP describes itself as “an open-source standard for connecting AI applications to external systems,” the standard other clients, Claude and ChatGPT among them, build support for independently. LangChain used to reach MCP servers through a separate langchain-mcp-adapters package; that’s now folded into LangChain itself, in a langchain.mcp namespace built on FastMCP, requiring langchain[mcp]>=1.4.0. It’s still marked beta: importing from it raises a warning, and the interface can still change.
MCP and LangChain aren’t really the same kind of thing to begin with: MCP is a protocol for where a tool’s definition and execution live, not a framework for building the agent around it. Adding LangChain’s MCP client doesn’t change what a tool call looks like to the model; it only changes where the tool it’s calling actually runs.