Organization repository on GitHub
Jev Browser — indexed browser automation. Jev chooses the control, Playwright acts. A CodexQA skill.
- Forks
- 6
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
CommunityJev Demos & Experiments
CodexQA Jev Browser
A Playwright-based CodexQA skill where Jev chooses indexed controls on a page-built index while the browser executes the action.
- Stars
- 97
- Language
- TypeScript
- License
- MIT
- Published by
- Community
- Author
- openqa-cn
- Last verified
- Oct 2026
About this project
简体中文 · How it works · Known limitations. Jev Browser is one skill in CodexQA: local Agent Skills that check whether code is actually good after it is written. This repository is the standalone project. The same skill is also installed from the CodexQA catalog. GUI-model browser automation sends a screenshot to a vision model on every step. Recognition spends vision tokens, the loop waits for the model to read the image, and the click lands on coordinates.
Quoted from the project's README. Read the full README.
- Languages
- Python
- Platforms
- PlaywrightCLI
- Capabilities
- Open sourceRequires API key
Highlights
- README: Jev chooses an indexed control; Playwright Chromium performs the click or typing.
- Knowledge-base markdown files supply domain facts Jev cannot infer from the DOM alone.
- Shares the CodexQA skill catalog pattern for post-write code quality checks.
Watch out
MIT. Requires curated knowledge/<app> notes for reliable flows; marketing latency figures cite TypeSafe launch posts, not this repo's benchmarks.
Questions about this entry
Short answers drawn from the linked project or page.
No. Observation runs inside the page and gives every visible control an index, a role, a name, and the operations it supports; the decision must return an index and an operation from that closed set. Selector-like text, JavaScript, and shell in the model reply are rejected before anything runs, and screenshots stay in the report only to mark the control that was used.
The knowledge base is the main lever. Jev can only pick an indexed control and can only type phrases already present in the goal, so it does not know that a cite is an ad or that a login dialog means stop. Those sentences belong in knowledge/<app>/*.md, and the README says a failed live run is usually a missing or wrong note, not a missing selector.
No. The repository states plainly that it has no benchmark against vision GUI models. The 40x-200x speed range and the workflow cost comparison come from TypeSafe's published System One material and cover the decision call only, not loading the page or writing the report.
Only for the live paths. observe, run, explore, and --decisions make no model call at all; auto and generate --goal add a planner plus a decision provider. With TYPESAFE_API_KEY unset, an OpenAI-compatible chat/completions call handles the whole decision and the done check. Node 20 or newer is the only stated requirement.
npm install, copy .env.example to .env, then commands such as npx codexqa-jev-browser observe <page>, run <case>, auto --url ... --goal ..., generate ... --verify, and explore .... Node.js 20 or newer is required.
With TYPESAFE_API_KEY each step is a /systemone questionnaire choosing the operation and observed target, judging the step effect, and checking whether the task is done. An OpenAI-compatible chat/completions call is the fallback decision model, and --decisions skips both.
Notes in knowledge/<app>/ are the main lever. Jev only chooses an indexed control, so facts like a cite being an ad or a login dialog meaning stop belong in the notes; a failed run is usually a missing or wrong note, not a missing selector.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Jev driving real Chrome from Grok Bot
X: jev-browser, a general browser skill
X: A second-hand shopping agent
X: A job crawler that reads careers pages with Jev
More Jev resources like this
29.3kGitHub stars
The Cua computer-use platform documents a jev-use recipe where TypeSafe Jev picks bounded action IDs and Cua Driver executes clicks and fills in the browser.
Jev Demos & Experiments#community#browser#computer-use
Community
trycua/cuaRust
trycua/cuaRust221GitHub stars
An agent skill that reproduces Jev-style single-step logits over indexed browser controls using local Qwen weights for offline browser use.
Jev Demos & Experiments#community#browser#open-models
Community
APUS-AI-Lab/fast-browser-usePython
APUS-AI-Lab/fast-browser-usePython634GitHub stars
A CLI browser automation tool with optional Jev semantic mode: Surf observes, Jev selects the next action, Surf validates and acts, Jev checks the goal.
Jev Tools & CLIs#community#browser#cli
Community
nicobailon/surf-cliTypeScript
nicobailon/surf-cliTypeScript306GitHub stars
Local macOS computer-use automation on Apple silicon using on-device laya typed decisions and CUA-S1-FORMS for filling, via the Accessibility API.
Jev Demos & Experiments#community#browser#local
Community
FluidInference/FluidUseSwift
FluidInference/FluidUseSwift36GitHub stars
A Clash Royale bot that turns a phone screenshot into JSON once a second and asks Jev for a strategy, a card, and a square.
Jev Demos & Experiments#community#python#gaming
Community
bytelabs-oss/clash-jevPython
bytelabs-oss/clash-jevPython11GitHub stars
Open-source, pre-alpha CLI that runs plain-English browser tests on Playwright, using Jev for bounded decisions: resolving each step, picking elements among offered candidates and judging verify claims with probabilities. Bring your own TypeSafe key; MIT licensed.
Jev Tools & CLIs#community#e2e-testing#playwright
Community
sedum-dev/sedumTypeScript
sedum-dev/sedumTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
Browser loops are long series of small decisions, and the expensive model does not need to make any of them. The patterns behind the fastest agents, from element tables to OCR plus accessibility trees.
Jev Tools & CLIs#browser#computer-use#agents
Read article
A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.
Jev Demos & Experiments#gaming#agents#control
Read article
From the Jev community
Posts from builders shipping with Jev right now.