Hello,
I see the industry uses Langchain a lot (job posts) but I’ve never seen MCPs in them.
Were MCPs thrown away as an unnecessary standard instead of standard tool calls?
Just curious because I can’t get this info in nowhere but from pracititioners.
Cheers.
I can’t predict future trends, but I think MCP is widely used. As the name suggests, MCP is a protocol, so it defines how requests and data are exchanged.
It exists to turn existing software into MCP servers or add MCP client functionality to existing software. Furthermore, frameworks already exist for this purpose, so aside from building the core software itself, it’s not difficult enough to warrant hiring…
The basic use case of MCP is turning AI or regular programs that handle simple tasks into MCP servers, making them the LLM’s hands and feet within the computer. Therefore, it’s unlikely to become a very large-scale project.
LangChain, on the other hand, is a single large framework. It’s essentially a collection of numerous software components, with its role being the bridge between them. LangChain can handle a wide range of tasks independently. While it operates within Python, it effectively functions like a programming language in its own right, requiring a learning curve. For long-term operation within an enterprise, it currently seems easier to build stable software with LangChain. This nature makes it more directly applicable to actual business operations…
BTW, MCP can be used from LangChain too…