Using MCP servers from ChatGPT
Warning: Following this article implies that you are going to expose local services of your computer openly to the Internet. If you aren’t familiar with running VPN’s and services (and if you do, there is a good chance you think I’m crazy for merely suggesting this setup) I do not recommend following it. I especially advise you against running tailscale on any of your work devices, unless you want to be fired. You have been warned. If you have been following AI trends you will have heard about MCP or the Model Context Protocol, and the sudden explosion of mcp servers (thousands of them, if we believe the newly created “MCP registries”). It’s one of those things that seem obvious in retrospect, but caught almost everyone by surprise. MCP is no more than a standardized way of letting LLM models access “tools”. These tools can be anything that can be accessed by a software, and includes reading and writing local files, accessing databases, browsers, remote API’...