AI assistant and MCP: manage your VPN in plain words

In the previous lessons we took the panel apart piece by piece: servers, routes, users, the storefront, and payments. You don't have to manage all of this with buttons alone — you can connect an AI assistant to your account and give it tasks in plain words.

For example:

Find the user by their email address and explain why their VPN isn't working.

Check the server in Frankfurt and briefly tell me whether everything is fine with it.

Create a "Germany" route from two ready-made inbounds.

The assistant doesn't guess the state of your account and doesn't click buttons in a browser. It receives the tools it's allowed to use through MCP and works with CreateYourVPN directly.

What MCP is

MCP (Model Context Protocol) is an open protocol that lets an AI app connect to external data and actions. It's easiest to think of it as a universal plug:

• AI model — reasons and decides which tool is needed: ChatGPT, Claude, DeepSeek, or another model that supports tool calling;
• MCP client — the app where the conversation happens and that can connect MCP servers;
• CreateYourVPN MCP server — describes the available actions and safely performs them in your account.

The model and the client aren't necessarily the same thing. For example, DeepSeek can work as the model inside Cherry Studio, and Cherry Studio will be the MCP client. The chat.deepseek.com website itself plays no part in this setup.

What the assistant can do

The set of tools depends on the access level you choose, but with the full setup the assistant can handle almost all of the everyday work in the panel.

Users and access

• find a user and show their status, expiry date, traffic, and devices;
• create a user, change limits, or grant access;
• get a subscription link and diagnose a client's problem;
• delete a user after an additional confirmation.

Servers and infrastructure

• show clusters, servers, inbounds, and routes;
• connect a new VPS and follow the installation stages;
• run server diagnostics over SSH and explain the problems it finds;
• create or change an inbound, masking, split tunneling, and a route;
• look through the history of recent incidents.

Storefront and account

• change the name, colors, contacts, and other storefront settings;
• work with plans and your own domains;
• show the balance, current charges, and the activity log;
• explain how CreateYourVPN works using the built-in help and Academy articles.

Three access levels

When you connect, CreateYourVPN shows a separate permission page. On it, you choose what the assistant will be able to do.

Start with the View level. If you later need an action that isn't in the granted set, disconnect the app and connect it again with a higher level.

How your account is protected

The connection works through OAuth: your CreateYourVPN password and your partner API key are never passed to the AI app. You sign in on a CreateYourVPN page, see the permissions being requested, and approve them yourself.

On top of that, four safeguards are in place:

1. Limited permissions. The assistant gets only the level you chose when connecting.
2. A code for deletions. Deleting a server, inbound, route, user, plan, or domain requires a six-digit code from an email sent to the account owner. The assistant first explains the consequences and only then asks for the code.
3. Activity log. The Connected AI → AI actions page shows which app called which tool, and when.
4. Instant disconnect. You can revoke access for a specific app at any time in the Connected AI section.

What happens after you send a message

Let's walk through the request "Why isn't this user's VPN working?" step by step.

1. The AI understands that it needs facts, not general advice.
2. The MCP client shows the model the CreateYourVPN tools available in this connection.
3. The model calls the user search and the diagnostics for that user's subscription.
4. The MCP server checks the OAuth access, runs the request inside your account, and returns a structured result.
5. The model turns the technical answer into plain language and suggests the next step.

CreateYourVPN remains the source of facts and the one that carries out actions, while the model is the conversation partner that picks the right tool and explains the result.

How to connect an AI app

The exact menu names differ from app to app, but the overall flow is the same.

If the app has no option for adding your own remote MCP server, you won't be able to connect CreateYourVPN through it. The model may still support tools while the client doesn't: both parts matter.

How to give instructions

A good task names the object, the result you want, and the limits:

• "Find the user user@example.com, check their subscription, and only explain the problem — don't change anything."
• "Show the servers that are offline, then run diagnostics on the most problematic one."
• "Suggest a route for Germany made of ready-made inbounds, but show me the plan before you create it."
• "Compare the daily charges with the balance and tell me how many days it will last."

For changes, it helps to ask for a plan first. And once the work is done, ask for a list of exactly what changed. The CreateYourVPN log remains an independent way to check what the assistant did.

Remember

• MCP connects an AI app to CreateYourVPN tools; it isn't a separate AI model.
• The model thinks and explains; the MCP server reads data and performs the allowed actions.
• It's safer to start with the View level.
• Deletions require a code from an email, every call goes into the log, and the connection can be revoked.
• Support depends not only on the model, but also on the app it runs in.

What to read next

Want to use DeepSeek specifically? In a separate guide, we'll install Cherry Studio, add a DeepSeek API key, connect the CreateYourVPN MCP server right from the chat, and check the first request.