What is MCP?

In one sentence

MCP, the Model Context Protocol, is an open standard that lets an AI agent connect to an outside service and use its tools, a bit like a universal plug between AI apps and the services they work with.

In more detail

Anthropic published MCP as an open standard in late 2024, and many AI apps now support it, including Claude, ChatGPT, Cursor and others.

A service that offers an MCP server describes its tools in a way any MCP-compatible AI can understand: for example "list conversations" or "send a message". The AI then decides when to use each tool.

For you, it means one connection works with several AI agents. You add the service once to the agent you prefer, and you can switch agents later without rebuilding anything.

Before a shared standard, each AI app needed its own custom connection to each service. A service that wanted to work with five AI apps had to build five different connections, and an owner who changed AI app often had to start again. MCP replaces that with one common way of describing tools.

MCP does not make an AI cleverer, and it does not give it permission to do anything on its own. It only describes which tools are available and how to use them. What those tools are, and what limits apply to them, is still decided by the service on the other end.

A simple way to picture it

Think of a travel adapter. Your AI app is the device, and each service, such as your calendar or your messaging accounts, is a socket in a different country. MCP is the adapter that lets the same device plug into any socket that supports it.

The socket still decides how much power comes through. In the same way, the service decides which tools your AI agent gets and how often it can use them. If a service caps how many messages an account can send in a day, the AI cannot change that, however it is asked.

Words you will see

What happens when your agent answers a message through MCP

1. A customer sends a message to your WhatsApp number, as they always have.
2. Your AI agent uses the Spaces tool for listing conversations and sees a new, unanswered message.
3. It uses the tool for reading messages to see the whole conversation, including any photos the customer sent.
4. It checks your notes, decides what to say and uses the reply tool.
5. Spaces checks the reply against the limits for that account: the gap since the last message, the daily total, and whether the customer has said stop.
6. If everything is fine, the reply goes out from your own number. If it is too soon, Spaces holds it and tells the agent how long to wait.
7. You see the reply in WhatsApp on your phone, in the same chat, as if you had typed it.

In your AI app you might see tool names such as list_messaging_chats, list_messaging_messages and reply_in_chat. You do not need to remember them, because the agent reads each tool's description and picks the right one itself.

How to add Spaces to your AI agent

Every AI app uses the same server address, https://api.spacesos.com/mcp, and signs in to Spaces in your browser. An API key also works for apps that prefer one. Spaces has a setup page for each of these apps, checked against that app's own instructions:

• Claude Code
• Claude
• Cursor
• Codex
• ChatGPT
• Grok
• OpenClaw
• Hermes

In Claude on the web or in the desktop app, the steps are:

1. Open Customize, then Connectors, then Add custom connector.
2. Name it Spaces and paste https://api.spacesos.com/mcp as the URL.
3. Choose Sign in, then sign in to Spaces and approve access.
4. On a Team or Enterprise plan, an owner adds it once under Organization settings, then Connectors.

In Claude Code it is one line, followed by a sign-in:

claude mcp add --transport http spaces https://api.spacesos.com/mcp

Menu names in AI apps change from time to time, so if a step looks different, the setup page for that app links to its own help page.

One server for everything, or one per channel

The main Spaces address covers every channel you connect. There are also addresses for a single channel, such as https://api.spacesos.com/mcp/whatsapp or https://api.spacesos.com/mcp/instagram. They use the same key and the same limits, but the agent only sees that channel's accounts and tools.

A single-channel address is useful when you want an agent that can only ever touch one channel, for example a WhatsApp agent for bookings that has no access to your LinkedIn.

MCP or the API: which do you need?

Most small businesses only need MCP. Spaces offers both, with the same limits, so a developer can build on the API later without changing anything about your accounts.

What MCP does not do

• It does not change any platform's rules. WhatsApp, Instagram, LinkedIn and the others still decide what is allowed on their services.
• It does not teach the agent about your business. Your notes still do that.
• It does not give the agent access to anything you have not connected.
• It does not remove the need to check the agent's work, especially in the first few weeks.

Keeping it safe

• Only connect services you trust, and only the ones your agent needs.
• Prefer the sign-in option where your AI app offers it, so there is no key to copy around.
• If you do use an API key, keep it private and give each AI app its own.
• Look at the list of tools a service offers before you rely on it.
• Remember you can remove a connector from your AI app at any time.

Why it matters for your business

• You are not locked in to one AI company.
• Setting up is usually one step: add the server address to your AI agent and sign in.
• The service decides which tools exist and what limits apply, so the AI cannot do more than you allow.
• It works with the AI app you may already pay for, so there is little new to learn.

How Spaces handles it

Spaces runs an MCP server, so an AI agent that supports MCP can read and reply to messages on the accounts you connect. Spaces keeps the pacing and daily limits on its side, whichever AI agent you use.

The same tools cover WhatsApp, Instagram, LinkedIn, Messenger, X and email, so one agent can answer customers wherever they write. If an agent asks for something over a limit, Spaces says no, sends nothing and tells the agent when it can try again.

In short

MCP is the shared plug that lets an AI agent use outside tools. For you, it means you can add Spaces to the AI app you already like, switch apps later, and know that the limits on your accounts stay in place whichever AI is doing the work.

If you are ready to try it, open the setup page for your AI app, add the Spaces address and sign in. Then ask your agent to list your connected accounts to check it worked.

Do I need to be technical to use MCP?

No. Most AI apps let you add an MCP server in their settings, and many connect with a simple sign-in. If you can paste a web address and sign in, you can set it up.

Is MCP only for Claude?

No. Anthropic started it, but it is an open standard and many AI apps support it, including ChatGPT, Grok and Cursor.

Is it safe?

You control which services your agent connects to, and each service limits what the agent can do. Keep any keys private, prefer sign-in where it is offered and remove connectors you no longer use.

Does MCP cost anything?

MCP itself is an open standard, so there is nothing to pay for the standard. You pay for the AI app you use and for the services you connect, such as Spaces.

Can I use more than one AI agent with Spaces?

Yes. Each one connects with its own sign-in or key. The limits belong to each connected account, so two agents working on the same number share that number's daily allowance.

What if my AI app does not support MCP?

A developer can use the Spaces REST API instead, which has the same tools and limits. Otherwise, the agents page lists the apps that work with Spaces today.

Can the agent do things I did not ask for?

It can only use the tools the server offers, within the limits the server sets. Your notes decide which of those it uses, so be clear about what it should and should not do.