Inspiration
We've all been there: 20 tabs open, notifications pulling you everywhere, and two hours gone with nothing done. Existing tools block sites from lists you configure manually. We wanted something smarter, a tool that understands what you are actually trying to do and protects that intention automatically.
What it does
TabMind is a Chrome extension that tracks tab switching in real time, computes a live distraction score, and nudges you when focus starts fragmenting. Shield Mode closes junk tabs, blocks distracting domains, and uses AI to generate a Goal: one specific task broken into 3 to 5 actionable missions to complete right now.
How we built it
Built on Chrome Extensions Manifest V3 using chrome.tabs and chrome.storage. The AI layer is a local FastAPI server calling the Gemini API across three endpoints:
/infer-goalclassifies open tabs and identifies your current goal/generate-questproduces the focused task and mission breakdown/summarize-tabsgenerates one-line summaries for rapid tab triage
The popup is vanilla HTML, CSS, and JS talking to both the background service worker and the backend.
Challenges we ran into
Manifest V3 service workers have no persistent state, so syncing session data requires careful use of chrome.storage. Getting Gemini to return reliable, structured JSON from unpredictable real-world tab sets took significant prompt iteration. Notification timing was also tricky: too frequent and the tool becomes a distraction.
Accomplishments that we're proud of
The quest generation works. Opening 15 chaotic tabs, hitting Shield Mode, and watching the AI hand you a ready-made mission list with no manual input is the moment the tool clicks. We kept scope tight and shipped end-to-end in 24 hours.
What we learned
Tab switching frequency is a strong proxy for cognitive load. If a user switches tabs more than \( n \) times in a 10-minute window, focus is likely compromised. Prompt engineering for consistent, structured output is its own discipline. Most importantly: scope ruthlessly, merge early, rehearse the demo.
What's next for TabMind
- Focus window prediction based on personal session history
- Daily distraction report
- Team mode for remote workers
- On-device inference so nothing ever leaves your machine
Built With
- gemini
- javascript
- manifest.json
- python
Log in or sign up for Devpost to join the conversation.