What is a Custom Port?

A port is like a doorway into your server. Different apps use different ports:

• Jupyter: Port 8888

• TensorBoard: Port 6006

By opening a custom port, you can view your app in the browser from your Inference.ai dashboard without complicated networking steps.

How to Launch a Server With a Custom Port

1. Navigate to the Servers tab on the left sidebar

2. Select Create New Server

3. In the Edit Configurations segment, select the Advanced Options dropdown and look for Custom Port

4. Enter a port number of your choice (between 1,000 and 99,999)

5. Select + to add the port

What are Exposed HTTP Ports?

Exposing HTTP ports allows your container to receive incoming web traffic on ports beyond the default one. When you expose a port, we automatically generate a public URL, making your service accessible from anywhere on the internet.

If you expose a port, you’ll receive a URL like your-service-8080.example.com . This URL routes directly to the specified port. This can be useful to:

• Hose multiple applications with different entry points

• Enable API endpoints that operate on separate ports

• Run services specifically on ports

Key Points About Exposed Ports

• Only one exposed port per server is allowed at a time

• Exposed ports are not publicly open on the internet—they’re tunneled securely through your Inference.ai dashboard

• Access is temporary and session-based—when the server stops, the link becomes inactive