What Is an MCP Connector? A Plain English Guide for Office Managers
An MCP connector is an address that lets an AI assistant such as Claude or ChatGPT look things up in another service directly, using an open standard called the Model Context Protocol. Before adding one, check who runs it, what it can see and whether it can change anything.
Reviewed on
What Is a Connector?
If you look after your company's office, you may have been asked to add a connector to Claude or ChatGPT, or seen a website invite you to. The word sounds technical, but the idea is simple. This guide explains what a connector is, what the letters MCP mean, and what to check before you add one at work.
A connector is a way for an AI assistant to ask another service a question directly and get a structured answer back. Without one, an assistant that wants to know about offices has to search the web, read pages and piece together an answer. With one, it can ask a marketplace for offices that match a request and receive the facts as data: the address, the number of desks, the price and so on.
What Does MCP Stand For?
The Model Context Protocol. Anthropic, the company behind Claude, released it as an open standard in November 2024, describing it as a standard for connecting AI assistants to the systems where data lives[1]. The protocol's own documentation compares it to a USB C port for AI applications[2]: just as one kind of plug connects many devices, one protocol lets many assistants connect to many services.
What Is an MCP Server?
It is the service on the other end of the plug. A company that wants assistants to be able to reach its information runs an MCP server at a web address. When you add a connector, you are telling your assistant that address. The assistant then decides when to use it, based on what you ask.
What Can a Connector Do?
That depends entirely on how it was built. Some can only look things up. Others can take actions on your behalf, such as sending a message, creating a document or booking something, and those usually ask you to sign in so they act as you. A read-only connector cannot change anything anywhere. For finding an office, a read-only connector is all you need.
What Does No Sign In Mean?
Some connectors need no account at all. Anthropic's documentation explains that with the no sign in option, anyone with access to the server's address can use the connector[3]. That is fine for public information, such as office listings, because there is nothing private to protect. It would not be fine for anything that reads your company's own data.
What Should an Office Manager Check Before Adding One?
Five things:
- Who publishes it: the company name, its website and how to contact it.
- What it can do: look things up only, or take actions as well.
- What it can see: only public information, or your files, email or calendar too.
- Whether it asks for sign in or personal details, and what its privacy notice says.
- What your organisation's policy allows.
On Claude's Team and Enterprise plans, Anthropic says only an owner, or on Enterprise someone with the right role, can add a custom connector for everyone[4], so your IT administrator may need to do it first.
How Do You Add One?
In Claude, you open the connectors settings, choose to add a custom connector, give it a name and paste its address. Anthropic says custom connectors are available on its Free, Pro, Max, Ultra, Team and Enterprise plans, with Free users limited to one[4]. ChatGPT uses a developer mode for custom MCP servers[5], and which plans can use it depends on OpenAI's current rules. Many services also publish their own step by step instructions.
What Does a Connector for Office Space Look Like in Practice?
District (district-hq.com), a managed office marketplace for London, publishes one, and it can be read against the five checks. District HQ Ltd, registered in England and Wales under company number 17386245, runs it. District offers a free, read-only connector that works with AI assistants such as Claude and ChatGPT. An assistant can search District's published, verified listings and open a single listing, and every result links back to the listing page on District. The connector cannot book anything, cannot see anyone's account and never shows an operator's contact details, so enquiries and viewings still happen on District. Its address and the steps to add it are on District's page for AI assistants. District's listings are being onboarded, and none are published yet.
What Happens After You Add It?
Nothing, until you ask. In a new chat, switch the connector on, then ask a normal question, such as an office for a team of ten in Shoreditch from January for a year. The assistant chooses whether to call the connector and usually tells you when it has. You can switch it off at any time, or remove it in the same settings where you added it.
A connector is just a direct line from your assistant to a source of facts. Treat it like any other tool your team uses: know who runs it and what it can reach, and you will get the benefit without the worry.
Sources
- Introducing the Model Context Protocol, Anthropic. Read on .
- What is the Model Context Protocol (MCP)?, Model Context Protocol. Read on .
- Add a connector that is not in the directory, Anthropic (Claude docs). Read on .
- Get started with custom connectors using remote MCP, Anthropic (Claude Help Center). Read on .
- Developer mode: connect a custom MCP server, OpenAI. Read on .