Unclutter
Classify distracting page elements once, then reuse conservative hiding rules for the same page template.
What Jev decides
Whether each candidate is useful content, an eligible clutter category, or uncertain.
The browser extension sends bounded descriptions of candidate elements to Jev, keeps only high-confidence clutter classifications, and stores local selectors that can be reapplied without another model request.
- Jev role
- Classify bounded DOM candidates so local policy can decide which nonessential elements may be hidden.
- Input
- Up to 60 candidate descriptions containing element type, structural signals, short redacted text, position, and match count.
- Output
- Typed Choice results with category probabilities for each candidate.
- Decision frequency
- On manual analysis, or once for a new page template when the user enables automatic analysis.
- Fallback
- Keep the element visible on low confidence, uncertainty, malformed output, unsafe selectors, or provider failure; existing rules remain unchanged.
How it works
Overview
Unclutter is a WXT extension for Chromium and Firefox. It extracts up to 60 bounded descriptions of page elements and asks Jev to classify each candidate as useful content, advertising, promotion, newsletter, social UI, a cookie surface, or uncertain. The extension validates returned categories and probabilities before creating reusable rules for that origin and page-template family.
Main content, navigation, forms, login, payment, security, and paywall surfaces are protected. Low-confidence or uncertain elements remain visible, hidden nodes can be restored, and saved templates work offline. The extension hides presentation only: it does not block requests, reject cookies, bypass access controls, or click consent buttons.
- Language
- TypeScript
- License
- MIT
- Author
- Kitze
Sources
Related projects
jev-ultrafast
A browser agent where Jev selects an operation and a compatible page element from a live indexed action space.
What Jev decides
Which supported browser operation to run, and which compatible visible element to target
typesafe-computer-use
A native macOS computer-use loop built from local OCR, accessibility data, and bounded Jev actions.
What Jev decides
Which computer action, visible item, site, or off-screen accessibility control to use next.
Jev Chat Assistant
An Android conversation copilot that uses Jev to interpret chats and rank reply candidates without sending for the user.
What Jev decides
The likely intent, conversation risk, reply need, best next action, unresolved tension, and which of three candidate replies fits best.