br-AI-n takes a local-first approach to browser AI. Instead of forwarding your prompts to a remote API, it downloads an open-weight model once into your browser cache and then performs inference in the sandboxed extension environment using WebAssembly and WebGPU hardware acceleration, with an optimized single-threaded SIMD path as a fallback on machines without usable GPU support. Because everything runs on-device, there are no per-token costs, no monthly fees and no need to paste in an API key.
The extension lives in Chrome's side panel and can be opened with a single click from any tab. Highlight text on a page and a floating quick-action widget offers instant summarization, concept explanations and grammar fixes without leaving the page, while an "attach page context" control pulls the current tab's text straight into the conversation for longer documents. Advanced reasoning models render their expandable thought steps natively in the interface.
For users who already run their own inference stack, br-AI-n can also act as a front end for local server backends such as Ollama, LM Studio and llama.cpp, offloading heavy work to a dedicated machine while keeping the Chrome interface. Optional web tools — search, active-tab reading and sandboxed code execution — are disabled by default and gated behind a fail-closed network kill switch, so the extension stays fully offline unless you explicitly opt in. Personas are customizable with preset profiles for assistant, coding, summarizing, translation and grammar checking, plus unlimited user-defined system prompts.
