# Connect

> Add the MCP server to Claude Code, claude.ai, Codex, Cursor or VS Code. About two minutes.

Add this URL to your agent app:

```text
https://mcp.trykubi.com
```

The first time, your browser opens our sign-in page. Pick the screens the agent may use and what it may do, then return to your app. What that grants: [Sign-in and permissions](/mcp/sign-in).

## Claude Code

```bash
claude mcp add --transport http kubi https://mcp.trykubi.com
```

Then run `/mcp` inside Claude Code and sign in (or `claude mcp login kubi` from your shell).

## claude.ai and Claude Desktop

Open **Customize > Connectors**, click **+ Add**, then **Add custom connector**. Name it Kubi, paste the URL above and click **Continue**, then **Add**.

## Codex

```bash
codex mcp add kubi --url https://mcp.trykubi.com
codex mcp login kubi
```

## Cursor

Add the server to `mcp.json`:

```json
{
  "mcpServers": {
    "kubi": { "url": "https://mcp.trykubi.com" }
  }
}
```

## VS Code

Open the Command Palette, run **MCP: Add Server**, choose **HTTP** and paste the URL.

## An API key instead of signing in

For agents that run without a person (CI, servers), send a key in the header:

```text
Authorization: Bearer kubi_...
```

The key's own permissions and screens apply. To name the runs in History, send `Kubi-Run: <a label>` too; without it each MCP session is one run.
