How I connected AI to my tools
The MCPs I use every day, how they're installed and the risk of giving it access to everything.
MCP (Model Context Protocol) is an open standard that lets you connect an AI to other tools and data: your email, your files, a database or the browser.…
MCP (Model Context Protocol) is an open standard that lets you connect an AI to other tools and data: your email, your files, a database or the browser. It was introduced by Anthropic, the company behind Claude, at the end of 2024, and today many other AIs and programs support it. In short: it's what lets AI stop talking and start touching your stuff.
The most useful comparison is a universal plug: before, each AI needed a custom-built connection for each tool; with MCP, all it takes is for both to speak the same language. That connection is made by an "MCP server", a small program that acts as a bridge to a specific tool and tells the AI what it can do with it. What confuses people at first is thinking MCP is an AI or a program you download and that's that: it's the way of connecting. And watch out: installing an MCP means granting permissions, so it's worth knowing where it comes from.
MCP appears in two parts of the blog. In "What surrounds the model to make it useful" I explain how I use it to connect my local models to my tools, along with other pieces that surround the model. And in "How I connected artificial intelligence to my tools" I go through the ones I use daily, how they're installed, what the risk is of giving it access to everything and which ones someone new should start with.
The MCPs I use every day, how they're installed and the risk of giving it access to everything.
What surrounds the model so it's actually useful: interfaces, RAG, connections and scripts.
One email a week with what was verified, what was refuted and what is still on the bench. No news roundups.