Connecting AI assistants to the real world with MCP

AI assistants are improving, not just at answering questions but at taking action. A big trend this week is building new tools for them using the Model Context Protocol (MCP).

I decided to experiment with this through my side project, isfree.se.

It's a simple tool that lets you instantly check the availability of a Swedish .SE domain by typing a URL. Quick and effortless, perfect for those moments when inspiration strikes and you need to see if your ideal domain is still available.

Now, thanks to MCP, an AI can do this check for you.

You can ask Claude to generate a company name for your dog skateboard startup (because why not?) and instantly check if the domain is available.

Is this really useful? Not very, but maybe a little.

It’s more of a fun prototype that shows how AI can interact with external systems.

Right now, MCP is still complex to configure, supported only by Claude among major AI assistants, and lacks a clear catalog of tools to browse. But with OpenAI recently announcing MCP support for ChatGPT, that could change quickly.

And once AI assistants can connect easily with external systems, things start to get really interesting.

But why stop there when you can try it yourself?