remote ollama access · openai-compatible
Keep Ollama on your own computer. 41tunnels gives it an OpenAI-compatible endpoint any app can reach, no router port to forward, no complicated setup, no usage bills.
// any OpenAI-compatible clientbase_url = https://relay.41tunnels.com/v1api_key = 41t_9f2c…model = llama3.1:8b // answered by the machine on your desk97 connections left on the current Relay server
It lives in the menu bar or tray and finds Ollama on localhost:11434.
Amallo hands you a base URL and a key: the OpenAI-compatible API is on out of the box. Paste them into whatever speaks it: Cursor, Aider, a script, a compose file.
The OpenAI-compatible API gives you a base URL and a key; any client that speaks it accepts them unchanged. A few people actually point at it:
No. Amallo makes one outbound connection and keeps it open, so nothing on your router changes.
Your computer does the work, so there is nothing to pay per message.
Same shape (expose a local port through a public relay), but scoped to Ollama rather than configured per tunnel. Amallo dials out on its own and hands you a ready OpenAI base URL, so there is no tunnel URL to protect and no account to wire up.
Yes. That is the intended setup for anyone who would rather not trust ours. The source is on GitHub, it builds to a single Go binary, and every release ships a container image; point Amallo at your own hostname.
Anything that speaks the OpenAI API, unchanged: editors, agents, CLI tools, your own scripts. A few examples ↓
v1.5.0 · macOS and Windows · all releases