What is an MCP server?
An MCP server is a small program that lets an AI app, such as Claude or ChatGPT, work with one piece of your business software. It tells the AI app what it may do there, and it carries out what the AI asks for. MCP stands for Model Context Protocol, a shared standard for connecting AI apps to other software.
Why it matters to a business
An AI app on its own knows nothing about your company. It cannot see your customers, your orders or your tasks. So people copy information into the chat by hand and copy the answer back out. An MCP server removes that copying. The AI app can look things up and make changes in the real software, within the limits you set.
Think of a waiter in a restaurant. The waiter hands you the menu, takes your order to the kitchen and brings back the dish. You never walk into the kitchen yourself. An MCP server does the same job between an AI app and your software.
What is on the menu
When an AI app connects, the server hands it a list of actions. Each action has a name and a short description in plain language, such as "find a contact", "create a task" or "read this file".
You then ask for something in your own words, for example "Add a task to call the supplier on Monday". The AI reads the list, picks the action that fits and asks the server to carry it out. One server serves one piece of software: your customer list, your file store or your accounting tool.
Where an MCP server runs
There are two kinds. A local server runs on your own computer and has to be installed there. It is mostly used by software developers. A remote server runs on the internet and is looked after by the company that makes the software. You install nothing, it is kept up to date for you, and the whole team uses the same address.
For a business, the remote kind is the one that matters.
How to connect to one, in four steps
- Find the address. The software that offers an MCP server shows a web address for it, usually on its settings page.
- Add it to your AI app. AI apps have a place for this, often called "connectors". You paste the address there.
- Show who you are. Either you sign in, or you paste a personal key. A key is a kind of password that the software creates just for you.
- Ask something. The AI picks an action, the server carries it out and the result appears in your chat. Many AI apps ask for your permission the first time an action is used.
That is all. No programming is needed to use an MCP server.
Four things to check before you connect
- Who made it? A server from the maker of the software is the safest choice. They know their own rules about who may see what.
- What can it change? It should be easy to see which actions only read information and which ones change it.
- Does the AI act as you? With a personal key, the AI can see only what you are allowed to see.
- How do you disconnect? There should be one obvious place to switch access off.
How Worknano handles this
Every Worknano workspace has its own remote MCP server. Its address is your workspace address with /mcp at the end. Each person connects with a personal key, so the AI app works as that person. Through the server, an AI app can read records, create and update records where an admin allows it, and read and save files. It can also propose changes to the setup as drafts that an admin approves. The server stays off until an admin turns it on.
- The MCP setup guide, with ready-to-paste examples
- MCP in Worknano: your AI apps working inside your workspace
Create a free workspace and see what its MCP server offers your AI app.