Are MCPs really useful?

Hello, hope everyone is well.

I’m quite new to Cursor and I’ve been hearing about MCPs non-stop. Does anyone have any experience with MCPs? In particular, are they useful? And is smithery AI safe?

Any recommendations? Or is it just a hype?

Moreover, I can’t get my codes to run on Cursor; they do run pretty well on jupyter, but even when I select Jupyter’s main Kernel as my Interpreter in Cursor, it doesn’t run. Do I need to install a specific library or something?

Appreciate your help.

Kindest regards, Masih

1 Like

MCPs are pretty useful if you want to extend Cursor’s capabilities - they let you add custom tools for reading/writing code, searching codebases, and connecting to external services. Check out Cursor – Model Context Protocol for the full details

Smithery runs MCP servers on your behalf, instead of your running them locally, which makes it quicker to get started with MCP. It’s not an official Cursor project, but should work in Cursor just fine!

Dear Dan, thank you for your comprehensive response!

Can I directly connect github for instance to Cursor without using a platform like Simthery?

My main concern is the safety of these new thirdparty tools like Smithery, and every tutorial I’ve read says use at your own risk!

Kindest regards, Masih

You can run MCP projects from Github locally. That way you can audit the code yourself.

May I ask how? And what do you mean by MCP projects? Isn’t MCP a server that we provide Cursor so it can do things on our behalf?

Hello again. Hope you are well; and thank you for responding to my earlier message.

When you said run MCP from github locally, did you mean to run git locally too? But will that update my repos in git in its public domain?