Connecting Vector MCP to your AI client

Last updated: May 8, 2026

This guide walks through connecting the Vector MCP to every major AI client. Setup is the same shape everywhere: paste a URL, authenticate once in the browser, done.

Server URL: https://mcp.vector.co/mcp

Auth: OAuth 2.0 (browser-based, no API keys to manage)

Time to set up: ~2 minutes

Before you start

You need:

  • An active Vector account. If you don't have one, ask your admin to invite you, or sign up at vector.co.

  • An AI client that supports remote MCP servers. The clients covered below all qualify.

You do not need an API key, a CLI, or any local installation.

Claude (Desktop or claude.ai)

Claude supports the Vector MCP as a Custom Connector on Pro, Max, Team, and Enterprise plans.

  1. Open Claude and go to Settings → Connectors.

  2. Click Add custom connector.

  3. Fill in:

    • Name: Vector

    • Remote MCP server URL: https://mcp.vector.co/mcp

  4. Click Add.

  5. Find the Vector tile in your Connectors list and click Connect. Your browser opens.

  6. Log in with your Vector account and grant access.

  7. You'll be redirected back to Claude with a "Successfully connected" confirmation.

Verify it's working. Start a new chat. Click Search and Tools in the composer and confirm Vector appears in the list with the tool count next to it and you should see several tools.

ChatGPT

ChatGPT supports the Vector MCP via custom connectors on Pro, Plus, Business, Enterprise, and Education plans. Free plans don't have access.

Heads up: On Business, Enterprise, and Education plans, you must be a workspace owner or admin to enable Developer Mode and create connectors.

  1. Open chatgpt.com on the web.

  2. Go to Settings → Apps & Connectors → Advanced settings (at the bottom of the page).

  3. Toggle Developer mode to ON.

  4. Go back to Settings → Apps & Connectors.

  5. Click Create. (This button only appears with Developer Mode on.)

  6. Fill in:

    • Connector name: Vector

    • Connector URL: https://mcp.vector.co/mcp

  7. Click Create. ChatGPT detects the OAuth requirement and opens your browser.

  8. Log in to Vector and grant access.

  9. Open a new conversation to start using the connector.

If ChatGPT ignores the connector, prompt it explicitly: "Use the Vector MCP to answer this — show me which companies visited our pricing page this week." Once it uses the tools once, it tends to keep using them.

Cursor

  1. In Cursor, open Settings → MCP.

  2. Click Add new MCP server.

  3. Use this configuration:

    {  
  "mcpServers": {
      "Vector": {
        "url": "https://mcp.vector.co/mcp"    
     }  
  }
}

  4. Save. Cursor detects the OAuth requirement and opens your browser.

  5. Log in to Vector and grant access.

  6. Return to Cursor. Vector tools are now available in chat.

VS Code (Copilot Chat)

VS Code supports remote MCP servers via Copilot Chat.

Option A — Command Palette

  1. Open the Command Palette: ⇧⌘P (macOS) or Ctrl+Shift+P (Windows/Linux).

  2. Run MCP: Add Server and choose HTTP.

  3. Enter the URL: https://mcp.vector.co/mcp

  4. Name it Vector MCP.

  5. Choose Global (user profile) or current workspace.

  6. VS Code opens your browser for OAuth. Log in to Vector and grant access.

Option B — mcp.json

Run MCP: Open User Configuration, or create .vscode/mcp.json in your workspace, and add:

{
  "servers": {
    "Vector MCP": {
      "type": "http",
      "url": "https://mcp.vector.co/mcp"
    }
  }
}

Save. VS Code detects the OAuth requirement and walks you through the login.

VS Code Insiders works the same way.

Claude Code

  1. Run claude to start Claude Code.

  2. Type /mcp.

  3. Select Add a server → enter:

    • Name: vector

    • URL: https://mcp.vector.co/mcp

  4. Select Authenticate. Your browser opens.

  5. Log in to Vector and grant access.

  6. Return to Claude Code. Confirm the server appears as connected with /mcp.

Codex

Codex auto-detects the OAuth requirement when you add the server.

  1. Add the Vector MCP to your Codex config:

    {  "mcpServers": {    "vector": {      "url": "https://mcp.vector.co/mcp"    }  }}

  2. Launch Codex with codex.

  3. Codex opens your browser for OAuth. Log in to Vector and grant access.

  4. Return to Codex and start using the tools.

Other MCP-compatible clients

Most clients accept this shape:

{
  "mcpServers": {
    "Vector": {
      "url": "https://mcp.vector.co/mcp"
    }
  }
}

When the client first connects, it detects the OAuth requirement. Some clients open the browser automatically; others require you to click a button. Log in to Vector, grant access, and you're done.

How OAuth works

The Vector MCP uses standard OAuth 2.0 with no client-side secrets. When your AI client connects:

  1. Your client detects that https://mcp.vector.co/mcp requires authentication.

  2. Your browser opens to the Vector login page.

  3. You log in (or are already logged in).

  4. You grant the MCP server access to your Vector data.

  5. Your client receives a scoped access token and stores it locally.

  6. All subsequent MCP requests use that token.

The token is scoped to your Vector account. The MCP can only access data you already have permission to view in the Vector UI. Nothing is shared across accounts. Tokens refresh automatically; if one expires, your client triggers a fresh OAuth flow.

If you ever need to revoke access, go to Vector account settings → Connected Apps and remove the entry for the AI client in question.

Troubleshooting

OAuth window doesn't open or hangs. Make sure pop-ups aren't blocked for your AI client and for vector.co. Close the tab and try Connect again.

"Successfully connected" but no tools appear. Restart your AI client. Some clients only refresh the tool list on startup.

ChatGPT doesn't show the Create button. Verify Developer Mode is on under Settings → Apps & Connectors → Advanced settings. Confirm your plan supports Developer Mode (Pro, Plus, Business, Enterprise, or Education).

Token errors after working previously. Remove the connector in your AI client and re-add it to clear stored tokens.

"Permission denied" on a query. The MCP only returns data your Vector account can access. Check your role in Vector — if you can't see the data in the Vector UI, the MCP can't either.

Stuck after these steps? Email support@vector.co with the AI client name, the exact error message, and a screenshot if possible.

Related