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