Connect to a GPU application using public web and direct ports

Before you start

Wait for the rental and application to be ready. Connection availability depends on provider networking and the endpoints you configured.

Steps

1. Open the GPU Instance page and its Connections section.
2. For a published HTTP application, use the displayed Open application link. Use the exact endpoint rather than guessing an IP or port.
3. For a direct public TCP or UDP endpoint, copy the displayed public address and mapped port into your client. Container ports and public ports can differ.
4. For an account-protected web endpoint, sign into the owning console account. For a template-protected application, use the rental’s application credentials.
5. If only SSH or a private endpoint is available, use the connection method shown or contact support about public-port availability. Do not publish an unauthenticated admin interface.

Check the result

The application loads or the client connects using the endpoint shown for that rental.

Troubleshooting

Check transport, HTTP versus HTTPS backend, bind address, process readiness, and application authentication. Not every provider offers direct public ports or private account protection; do not reuse an endpoint after the rental is destroyed.