MCP Server

The AI Admin Panel includes a built-in MCP (Model Context Protocol) server. This lets AI assistants like Claude Code, Claude Desktop, Cursor, and Windsurf manage your panel programmatically.

The MCP server is served by the panel itself — it shares the panel's domain and TLS certificate, so there is no extra port, subdomain, or firewall rule to open.

Enabling the MCP Server

The MCP server is disabled by default for security (its tools are admin-scoped). Enable it from the UI — no terminal or restart needed:

Go to Settings → MCP Server.
Toggle Enable MCP Server on.

The connection endpoints and ready-to-paste client configs appear once it's enabled. Disabling the toggle takes effect immediately.

Authentication

All MCP requests require an admin API key sent as a Bearer token:

Create an admin API key under Settings → Security → API Keys.
Use it as Authorization: Bearer <key> in your MCP client config.

Non-admin keys are rejected.

Endpoints

Both transports are served from your panel's domain:

Transport	URL	Use with
Streamable HTTP	`https://<your-panel>/mcp`	Claude Code, Cursor, modern clients
SSE	`https://<your-panel>/sse`	Claude Desktop, legacy clients

The Settings → MCP Server tab shows the exact URLs for your deployment.

Connecting from Claude Code / Cursor

{
  "mcpServers": {
    "ai-admin-panel": {
      "url": "https://your-panel.example.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_ADMIN_API_KEY"
      }
    }
  }
}

Connecting from Claude Desktop

{
  "mcpServers": {
    "ai-admin-panel": {
      "url": "https://your-panel.example.com/sse",
      "headers": {
        "Authorization": "Bearer YOUR_ADMIN_API_KEY"
      }
    }
  }
}

Available Tools

Tool	Description
`list_services`	List all deployed services with status
`get_service_status`	Get detailed service info including containers
`restart_service`	Restart a service
`stop_service`	Stop a running service
`start_service`	Start a stopped service
`list_templates`	Browse the template catalog
`list_customers`	List customers with plans
`get_dashboard`	Get dashboard metrics summary
`deploy_template`	Deploy a template with parameters
`get_logs`	Get service container logs

Example Usage

Once connected, you can ask your AI assistant:

"List all running services on the panel"
"Restart the n8n service"
"What templates are available?"
"Show me the dashboard metrics"