Skip to content
AI

MCP (Model Context Protocol)

MCP is a protocol that connects an AI tool such as ChatGPT or Claude to the data in your workspace, so the tool can look at it within the access you grant.

MCP (Model Context Protocol) is a shared way for an AI tool to reach data and actions in another system. Instead of building a separate connection for every tool, both sides follow the same rules. For a store owner the meaning is simple: from an AI tool you already use, you can ask questions about your customers and orders without opening another panel.

What the connection sets up is a permission. The tool sees and does only as much as you define when you connect it. Access can be kept narrow or widened. So the first decision when setting up MCP is what the tool is allowed to touch. When the data is about your customers, that decision matters more than how clever the tool is.

Why It Matters

  • You do not open a separate panel to look at data; you type the question into the AI tool you already use.
  • You choose the access level yourself, so what the tool can touch in your store is clear from the start.
  • MCP is a standard protocol, so the same connection logic applies across different AI tools.

How It Works

  1. Pick the toolChoose an AI tool that supports MCP. ChatGPT, Claude, Cursor and Windsurf are some of them.
  2. Connect the workspaceYou connect the tool to your workspace. Once connected, it can put questions to the data in that workspace.
  3. Choose the access levelYou decide how much access the tool gets when you connect it. There are three levels: read-only, can open drafts, and draft plus approval from chat.
  4. Ask in plain languageQuestions like “Who left a cart this week?” are written as normal sentences, and the tool fetches the answer from the connected data.

Example

Selin Kaya connects Liva Butik's workspace to Claude at the read-only level. On Monday morning she types, “Give me the list of customers who abandoned a cart this week.” The tool returns the list, and Elif Arslan's ₺890 Straw bag cart is on it. Selin reads the list and writes a note to pass to Burak Demir. Because the level is read-only, the tool changes no record and sends no message to anyone. Selin only wanted to look, and the connection allows exactly that much.

Not to Be Confused With

MCP is not an AI model or a separate product; it is the route a model uses to reach outside data. The MCP connection in Creato is also separate from AI Employees: AI Employees talk to customers, while MCP connects an AI tool you use to your workspace.

In Creato

Creato's external AI connection (MCP) is part of the paid plans.

  • You connect your workspace to supported AI tools such as ChatGPT, Claude, Cursor and Windsurf.
  • When you connect, you choose one of three approval levels: read-only, can open drafts, or draft plus approval from chat.
  • From the connected tool you can ask for the customer list and abandoned carts; those queries are read-only.
Explore MCP

Frequently Asked Questions

Do I need to know software to use MCP?

Daily use means typing questions in plain language. Setup means picking your AI tool, connecting your workspace and marking an access level. The exact steps differ a little by tool, so the description on the MCP product page is the best guide. If you get stuck, you can write to the support team.

How is data transferred over MCP, and which terms apply?

The access level you pick decides which data the tool can reach. The terms for data transfer are set out in a separate legal text: MCP Kullanım ve Veri Aktarım Koşulları. Read it before you connect, and choose the level with it in mind.

Can an AI tool send messages to my customers through MCP?

That depends on the level you choose. At the read-only level the tool just looks at data. Talking to customers is a different job, and AI Employees do it. MCP opens your own tool to your workspace; it does not put an employee in front of the customer.

Related Terms