Skip to main content

Dashboard

The Load Balancer provides a Dashboard for real-time monitoring and management of your Worker cluster.

Diagram

Features​

  • Lists all connected Workers and their status.
  • Displays Worker details (group, contract, slots, services, tags).
  • Shows logs from each Worker (fetched through the Load Balancer from their /logs endpoint).

Access​

  • The dashboard is available at /dashboard if enabled in the configuration.
  • The dashboard has no built-in login. Its page and its /dashboard/api/* routes, which return Worker URLs, Worker status and Worker logs, are accessible to anyone who can reach the Load Balancer. Protect /dashboard and all its subpaths with authentication and SSL at an HTTP proxy configured by the administrator, or keep the dashboard disabled.

Example Nginx Proxy Configuration​

location ~* ^/dashboard(/|$) {
proxy_pass http://localhost:4000;
auth_basic "Restricted";
auth_basic_user_file /etc/nginx/.htpasswd;
# SSL configuration here
}

Monitoring Workers​

  • The dashboard shows the Worker information the Load Balancer collects from each Worker's /health endpoint and WebSocket, and fetches logs from each Worker's /logs endpoint on demand.
  • Tags and other metadata are displayed for each Worker.
  • Use the dashboard to monitor cluster health, task queues, and debug issues in real time.

Enabling or Disabling the Dashboard​

The dashboard interface can be enabled or disabled using the DASHBOARD_ENABLED environment variable in your Load Balancer configuration. Set it to true to enable or false to disable the dashboard:

DASHBOARD_ENABLED=true

The dashboard is disabled by default. The sample .env enables it for local use; enable it in production only behind the authenticated reverse proxy described above.

MCP agent access​

The Agent access · MCP panel displays:

  • Whether MCP is enabled, its endpoint URL and a copy button.
  • The Streamable HTTP transport and synchronous execution mode.
  • Exposed tool titles and current Worker availability.

The panel does not ask for or store a token. It intentionally exposes only this limited operational summary. Use authenticated GET /health with LB_AUTH_KEY for target selections, schemas, reload timestamps and sanitized configuration errors.

MCP credentials and Worker keys are never displayed. Tools removed by a live configuration change disappear on the next refresh.

The existing reverse-proxy protection for the overall dashboard remains necessary. See LB MCP setup.