The idea is simple: if you can use a model from a browser chat, OpenBridge exposes that access as a local endpoint, so tools like OpenCode, OpenClaw, PI, or anything else that speaks the OpenAI format can use it too.

How it works:

* runs locally

* uses your existing authenticated web session

* translates requests and responses into a standard OpenAI-compatible interface

* supports normal chat flows and tool-style interactions

I’ve already used it to build multiple complex apps with it, including through OpenClaw, and it’s been great.

---

The motivation is bigger than just saving tokens.

I think inference should be as accessible as possible. If a person already has access to a model, they should be able to use that access from their own tools, on their own machine, without being forced into a separate paid API path just to automate legitimate personal workflows.

If your main concern is protecting the ToS of billion-dollar AI companies that ingested the open web at massive scale and now charge users for access to models trained on it, then this project is not for you.

https://github.com/uncensoredcode/openbridge

And yeah, this post was also written using OpenBridge.

Cheers, Linuz