How to enable the ExpressVPN MCP server on your AI tools

ExpressVPN just made something pretty cool — you can now hook your VPN directly into AI developer tools like Claude Code, Codex, Cursor, VS Code, and more, all thanks to the Model Context Protocol (MCP). No more alt-tabbing to the VPN app. Your AI assistant handles it.

Here’s everything you need to know to get it running.

What is the ExpressVPN MCP server?

The ExpressVPN MCP server (currently in beta) lets your AI coding tools — think Claude Code, Codex, Cursor, VS Code, or LM Studio — read and control your VPN connection using natural language or programmatic commands. We’re talking connecting/disconnecting, switching server regions, changing protocols, running speed tests, and even toggling ad/malware blocking — all without ever opening the VPN app itself.

It’s built on the Model Context Protocol, an open standard that allows AI assistants to interact with external tools and services. Anthropic introduced MCP in late 2024, and it’s been gaining rapid adoption across the developer ecosystem.

Related /

Before you begin: prerequisites

You’ll need a few things checked off first:

An active ExpressVPN subscription (1-year or 2-year plan on Basic, Advanced, or Pro tier — or ExpressVPN for Teams)
The latest ExpressVPN beta desktop app (macOS, Windows, or Linux)
An MCP-compatible AI client that supports HTTP-based MCP servers

Step 1: Enable MCP in ExpressVPN

Open the ExpressVPN desktop app
Go to Profile → Settings
Toggle on “Enable MCP Server”

That’s it. The local MCP server starts automatically in the background. It runs on 127.0.0.1:20090 by default — but check the app, as your machine may use a different port.

Step 2: Connect your AI tool

Pick your tool below and follow the one-time setup:

Claude Code

Run this in your terminal:

claude mcp add --transport http expressvpn http://127.0.0.1:20090/mcp

Codex

Add the following to your Codex MCP configuration file:

{
  "mcpServers": {
    "expressvpn": {
      "url": "http://127.0.0.1:20090/mcp"
    }
  }
}

Cursor

Go to Settings → Tools & MCP → + Add custom MCP, then paste:

{
  "mcpServers": {
    "expressvpn": {
      "url": "http://127.0.0.1:20090/mcp",
      "headers": {}
    }
  }
}

VS Code

Open the Command Palette (Shift+Cmd+P), run MCP: Add Server, and use:

{
  "servers": {
    "expressvpn": {
      "type": "http",
      "url": "http://127.0.0.1:20090/mcp"
    }
  }
}

LM Studio

Go to the right sidebar → Integrations tab → Install → Edit mcp.json and add the same config as Cursor above.

Any other tool

If your tool isn’t listed, just manually add http://127.0.0.1:20090/mcp it as an HTTP-based MCP server.

What can your AI actually do with it?

Once connected, your AI assistant can:

Connect/disconnect the VPN
Switch regions (e.g., “connect to Germany before running integration tests”)
Check your real and VPN IP addresses
Change VPN protocols (Lightway, OpenVPN, WireGuard)
Enable Network Lock (kill switch)
Set Split Tunnel rules per app
Toggle Threat Manager — ad blocking, tracker blocking, malicious site blocking
Run speed tests and diagnostics

It’s genuinely useful for developers doing geo-based testing or QA work across regions.

Is it safe?

Yes, and ExpressVPN is pretty upfront about the design:

The server is local only — it never talks to the internet, only your own machine
Your account credentials are never passed through the MCP interface
It’s opt-in — disabled by default
You can turn it off instantly by toggling MCP off in Settings

Quick troubleshooting

Can’t connect? Make sure the ExpressVPN app is open and the toggle is on. Run curl http://127.0.0.1:20090/mcp in your terminal to verify the server is live.
Wrong port? The app shows the actual port when you enable MCP — use that instead of 20090 if it differs.
AI using curl instead of MCP tools? Restart the AI client and re-run the MCP add command.
Firewall blocking it? On Windows, add an inbound rule for TCP port 20090. On macOS, check Privacy & Security → Local Network and grant permission to your terminal app.

The ExpressVPN MCP server is a neat step toward VPN management becoming a native part of AI-assisted developer workflows. If you’re already living inside tools like Cursor or Claude Code, it’s worth the five-minute setup.