![](/static/253f0d9b/assets/icons/icon-96x96.png)
![](https://beehaw.org/pictrs/image/1be75b15-2f18-429d-acf7-dcea8e512a4b.png)
Here is a possible case study for you. https://OpenRouter.ai.
They are doing what you think of doing, for the various LLM APIs. Three reasons why they offer added value:
- Some people cannot access some of these APIs directly (GPT-4 in the advanced settings, Anthropic’s Claude is still close access)
- They abstract the way these API are accessed so that the code to use different models does not have to change much.
- For some reason, they cost the same as the original API
There are some managed tunnel alternatives but they are not cheap. At this point it’s cheaper to rent a $20/year VPS and install one of these: https://github.com/anderspitman/awesome-tunneling
I personally love frp
Edit: I have never tried it, but there’s also Tailscale Funnel https://tailscale.com/kb/1223/tailscale-funnel/