The Model Context Protocol is an open standard for connecting AI assistants to tools and data. One server can be used by any assistant that supports it.

Is it safe to connect AI to company data?

It is when the access is narrow. I give each server only the tools it needs, keep it read-only where possible, and put an approval step in front of anything that changes data.

Which assistants does it work with?

Claude, ChatGPT and a growing list of agents and IDEs support MCP. I build to the standard, so you are not tied to one vendor.

Français

MCP servers: connect AI to your own tools and data

An AI assistant is only as useful as what it can reach. The Model Context Protocol (MCP) is the standard way to connect assistants like Claude and ChatGPT to your own systems. I build MCP servers that give them exactly the access they need, and nothing more.

What it looks like

Let your team ask questions of your own documents, CRM or database
Connect an assistant to an internal tool that has an API
Give an agent read-only access to records it should never change
Search your archive word for word, so answers quote the original
Control hardware, like a camera, from a chat

How it works

01Map the accessWe list what the assistant should be able to read and do, and what it must never touch.
02Build smallEach server gets a few clear tools. Read-only where that matters, with an approval step for anything that changes data.
03Test on real questionsI test it against the questions your team actually asks, and check the answers come from your records.

Things I have built

Dbrain
My self-hosted memory: records stored word for word and served to AI assistants over MCP.
Lightbox
An MCP server that lets an assistant browse my photo library.
Lodestar
It started as an MCP server with eight tools that drove a Sony camera over USB.
Emineon OS
A recruiting tool with a read-only MCP server.

Good to know

Works with Claude, ChatGPT and other assistants and agents that support MCP.
Servers can run in Switzerland, on your own hardware or behind a signed-in tunnel.
I work in French, English, German and Dutch.

Questions

What is MCP?: The Model Context Protocol is an open standard for connecting AI assistants to tools and data. One server can be used by any assistant that supports it.
Is it safe to connect AI to company data?: It is when the access is narrow. I give each server only the tools it needs, keep it read-only where possible, and put an approval step in front of anything that changes data.
Which assistants does it work with?: Claude, ChatGPT and a growing list of agents and IDEs support MCP. I build to the standard, so you are not tied to one vendor.

Tell me what you are working on

A few lines are enough. I reply by email. david.v@emineon.com