inbrowser.chat
Private AI chat that runs fully on-device in Chrome
Start new thread
Bas Hennekam

inbrowser.chat - Private AI chat that runs fully on-device in Chrome

by•
inbrowser.chat is a private AI chat that runs entirely on your device. It uses Chrome's built-in LanguageModel API (Gemini Nano), so every message is processed locally. No server, no API key, no network request. Chats stay in localStorage and work offline. Free, with no monetization planned. It's a demo of what a local LLM can do. Note by Google: this LLM is for summarizing, classifying or rephrasing text, NOT for use cases needing factual accuracy like answering knowledge questions.

Add a comment

Replies

Best
Bas Hennekam
Maker
📌
Hey Product Hunt 👋 I'm Bas, the maker of inbrowser.chat. Almost every "AI chat" sends your messages to a server somewhere. I wanted the opposite: a chat where nothing leaves your machine. inbrowser.chat talks to Chrome's built-in Gemini Nano model, so every message is processed locally by the browser itself...no server, no API key, no network request. You can open your Network tab and watch it stay empty. A few things I'm proud of: Truly private: conversations live in localStorage and never get uploaded. Works offline: once the model is downloaded. One single HTML file: hand-written CSS and vanilla JS, no build step, no dependencies. It needs Chrome 127+ with the built-in AI model enabled. I'd love your feedback on what on-device AI should do next: better models, file context, anything. Ask me anything!
Paul Jansen

How is this different from running Ollama or LM Studio locally?

Bas Hennekam
Maker

@mr_jansen Great question! Ollama and LM Studio are fantastic, but they're separate apps you install, and you have to download multi-gigabyte model files yourself and manage them.

inbrowser.chat takes a different angle: there's nothing to install at all. The model (Gemini Nano) is already built into Chrome, so the browser handles the download and runs inference through its native LanguageModel API. You just open a webpage and start chatting.

The trade-offs are honest:

  • Ollama/LM Studio let you pick and run much larger, more capable models. inbrowser.chat is limited to the small on-device model Chrome ships, so it's best for summarizing, rephrasing, and classifying rather than heavy reasoning or factual Q&A.

  • But you get zero setup, zero dependencies, it works offline once the model is cached, and the whole thing is a single hand-written HTML file. It's meant as a "look how far the browser alone can take you" demo more than a replacement for a full local stack.

So: think of it as the lightest possible way to try on-device AI, not a power-user toolkit. 🙂