Call your Ollama models from anywhere.

Set it up in four steps.

1. 1

   Start Ollama

   Run the server.

   ollama serve

   In a second terminal, pull a model.

   ollama pull llama3.2

2. 2

   Give it a private URL

   An internal endpoint is reachable from your ngrok account only. Nothing on your machine opens to the internet.

   ngrok http 11434 --url https://ollama.internal

3. 3

   Add it as a provider

   In app.ngrok.ai, create a custom provider with these values, then allow it on your access key.

   Provider ID

   ollama

   Base URL

   https://ollama.internal

   Model

   llama3.2

4. 4

   Call it

   Use any OpenAI-compatible SDK. Add a hosted model as a fallback for when your machine can’t answer.

   1import OpenAI from "openai";2 3const client = new OpenAI({⋯4	baseURL: "https://gateway.ngrok.ai/v1",5	apiKey: process.env.NGROK_AI_ACCESS_KEY,6});7 8const completion = await client.chat.completions.create({⋯9	model: "ollama:llama3.2",10	// If your machine is off or overloaded, try a hosted model next11	models: ["openai:gpt-5.4"],12	messages: [{ role: "user", content: "Hello!" }],13});