GitHub · Browser agents · Tools and apps

wy-coliney/jev-browser-use

The cheapest way to see the drive-with-Jev, verify-with-an-LLM split in a browser.

What it does

A browser skill that splits the loop in two: Jev owns clicking, toggling, scrolling and navigation, while Codex keeps text input, visual judgment and the final verification. The README claims 5–10x faster browser operations in the author's own product workflows at EZCollegeApp.

The repository's own description: “5–10x faster browser operations: Jev clicks, Codex thinks and verifies. Built at EZCollegeApp.”

How it works

Shipped as a skill rather than a library — plugin.json plus a skills/ directory and an INSTALL.md — so it attaches to a browser connection that already exists instead of bundling a driver. That is what lets the README promise no extra driver and no npm dependencies.

The repository, by the numbers

Stars

1,021

Licence

MIT

Stack

JavaScript, shipped as an agent skill (plugin.json + skills/)

Last push

2026-09-23

Read from the GitHub API on 2026-10-09. Stars and the last push move daily — quote them with the date, the way we do.

What we checked

the README: the "Jev clicks. Codex thinks and verifies." split, the 5–10x claim scoped to EZCollegeApp, and the no-extra-driver line
the repository layout: plugin.json, skills/, INSTALL.md, scripts/, assets/
GitHub metadata on 2026-10-09: MIT, 1,024 stars, last push 2026-09-23

What we did not check. We did not install the skill or measure anything; the speed range is the author's own workflow measurement.

Open the repository ↗All builds

The rest of the topic

Jev, the reference →

Start here

Jev, the reference What it is, what it costs, what we measured.
Every Jev guide, by question The whole topic as a contents page: what it is, how to build with it, what it gets used for.
How people actually use it Eleven projects read line by line.

Columns

The topic

The catalogue report What we counted and what we measured, with the JSON.
Jev with Claude Code Compaction, routing and reranking inside a coding session — and the job Jev must not take.

By use case

Browser agents 22 Driving a browser: clicking, reading, deciding the next step.
Triage and routing 47 Picking a route, a label or a next model — classification where a wrong answer is recoverable.
Coding and context 25 Inside a coding agent: compaction, context windows, the harness around the model.
Research and data 74 Benchmarks people ran, architectures people reverse-engineered, results that did not work.
Media and content 24 Generating, sorting or reviewing words, images and posts.
Trading and markets 7 Decisions with money attached, on a clock.
Games and real time 6 Playing a game, driving a live scene, or making a decision on a clock too short to think in.
Robotics and devices 3 Where the decision runs on the machine: on-device, local runtimes, simulated fleets, hardware in the loop.
UI 3 Where a person sees the result: design tools, interfaces that react while they are used, tagging and layout.
Tools and apps 13 Packaged as something you install or run: CLIs, browser extensions, plugins, database extensions, desktop apps.

What it does

How it works

The repository, by the numbers

What we checked

More like this

browser-use/jev-ultrafast

dzhng/jevgrep

szczyglis-dev/py-gpt