Lazy Cat

Inspiration

Hands-free browsing isn’t just a convenience it’s independence. Lazy Cat was inspired by the idea that voice-driven control can make the web easier for everyone, from multitaskers to people with limited mobility. I wanted to build something that lets anyone use Chrome comfortably without touching the keyboard or mouse.

What it does

Lazy Cat is a smart, on-device AI companion for Chrome. It listens for natural voice commands like “scroll down,” “open YouTube,” or “summarize this page,” and performs those actions directly in the browser. It can read, rewrite, summarize, translate, and proofread text all while staying entirely on your device.

Lazy Cat can also open websites, switch tabs, fill text boxes, click buttons, and even skip YouTube ads automatically when you say the word. It understands flexible phrasing so whether you say “scroll down,” “move down,” “slide lower,” or “jump to bottom,” it knows exactly what you mean.

And it can do even more

How we built it

Chrome Extension Manifest V3 (MV3)

JavaScript (background.js, content.js, popup.js, widget.js)

HTML & CSS (popup.html, persistent-popup.html, widget UI)

Chrome Extension APIs: chrome.runtime, chrome.storage, chrome.tabs, chrome.scripting, etc.

Chrome Built-in AI: Prompt API, Summarizer API, and Rewriter API (Gemini Nano).

The background script handles communication between the AI and the visible widget. The floating widget stays active across tabs, listens for “Hey Cat” or “Lazy Cat,” and executes actions on the active page.

Challenges we ran into

Keeping the widget alive between tab changes and making it aware of which page the user is currently on was one of the hardest parts. Manifest V3 removed persistent background pages, so maintaining state and re-injecting the widget consistently required a lot of debugging. Another challenge was making the voice commands feel natural people say the same thing in many ways, and the AI had to interpret them correctly without confusion.

Accomplishments that we're proud of

Lazy Cat feels like a real companion it doesn’t just listen; it understands. It recognizes flexible phrasing, interacts directly with web content, and even pastes rewritten or summarized text exactly where it came from. I personally use it every day just to skip YouTube ads hands-free it’s one of those small wins that make it genuinely useful.

What we learned

This was my first time integrating on-device AI into a real-world Chrome extension. I learned that AI is a bit like a three-year-old toddler who knows everything but still needs guidance to use it wisely. Understanding how to structure prompts, process responses, and safely handle context without storing data was a huge learning curve.

What's next for Lazy Cat

Next, I plan to:

Improve multi-language support with the Translator API.

Add offline TTS responses for smoother voice feedback.

Enhance context awareness so Lazy Cat can handle multi-step tasks.

Expand accessibility settings to support users with limited mobility or vision.

Built With

• chrome
• chrome.runtime
• chrome.storage
• css
• extension
• html
• javascript
• manifest