User repository on GitHub
⚡ 快速浏览器 Skill：Jev 驱动，Codex 即用。连续执行、预置输入，一句话启动浏览器任务。
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Agents & MCP
Jev Browser Skill — wanghai673
Codex Agent Skill that hands continuous browser work to Jev: one sentence starts a task, Jev picks the actions and targets, and Codex replies once the run completes. Reuses your own Chrome and reads the DOM instead of screenshots.
- Stars
- 2
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- wanghai673
- Last verified
- Oct 2026
This is one of several Jev listings for the same project. The canonical entry is Jev Browser Skill — hqman.
- Languages
- Python
- Capabilities
- Open sourceRequires API key
Highlights
- Turns one sentence into a browser task: Codex pre-supplies text like search terms, and Jev executes clicks, scrolling and page changes continuously.
- Reads page text and controls from the DOM by default, avoiding step-by-step screenshots and image understanding.
- README cites TypeSafe figures checked 2026-09-22: 70-500 ms model request latency, $0.042 per million input tokens, output free.
- The 7.1-second flight-search demo is attributed to the upstream browser-use/jev-ultrafast project, played at 1x and excluding initial page preparation.
- Supports macOS with Chrome and needs Python 3.12+ and uv; install by asking Codex to add the skill from the repository URL.
Quickstart
bash
python3 "${CODEX_HOME:-$HOME/.codex}/skills/jev-browser/scripts/jev.py" run \
--url 'https://www.bilibili.com' \
--goal 'Search Bilibili for machine learning and show the search results.' \
--inputs '[{"value":"machine learning","purpose":"Bilibili search query"}]' \
--showWatch out
Copy reviewed against the repo README on 2026-10-15; the README is in Chinese and the demo video, timings and latency figures come from the upstream browser-use/jev-ultrafast project. The README notes full tasks also include Codex, page loading and retry overhead beyond the per-request model figures.
Questions about this entry
Short answers drawn from the linked project or page.
A Codex Agent Skill that lets Codex operate a browser through Jev: you state a goal, Jev chooses actions and targets, and the skill runs consecutive operations in one go instead of handing control back at each step.
Send Codex the sentence in the README: install the jev-browser skill from https://github.com/wanghai673/jev-browser-skill. Setup then needs Python 3.12+ and uv, and the initial configuration is guided by the skill.
After installation you can say 'open Bilibili, search machine learning and show the results', or call the Python entry directly with --url, --goal, --inputs and --show flags, as shown in the README.
macOS with Chrome. The skill reuses your existing Chrome, and it needs Python 3.12 or newer plus uv; the README does not document Windows or Linux support.
It adapts the browser-use/jev-ultrafast runtime into a Codex Skill, adding pre-supplied inputs, continuous execution, configuration self-checks and result recording; the demo and performance figures come from that upstream project.
It reads page text and controls from the DOM instead of taking screenshots for image understanding, pre-fills text that Codex already supplied, and lets Jev chain clicks and scrolling without repeated model handoffs, per the README's cost table.
Yes, it uses the TypeSafe decision API, so a key is required; the README cites TypeSafe's figures of $0.042 per million input tokens with free output, checked 2026-09-22, but your full task cost also includes Codex and page-loading overhead.
More Jev resources like this
A research repository exploring which LLM-decision tasks Jev can accelerate, starting with a Codex plugin: Codex understands the goal and prepares inputs, Jev picks the next browser action, and the plugin executes it in local Chrome.
Jev Agents & MCP#community#catalog-import#logicrw
Community
tongyun1/jev-in-the-loopPython
tongyun1/jev-in-the-loopPythonA browser and macOS desktop agent for pi: Jev picks each action from structured observations of visible text and addressable controls — never screenshots — inside a bounded loop that stops when stuck or hands control back to a human.
Jev Agents & MCP#community#catalog-import#logicrw
Community
laihenyi/pi-jev-browserTypeScript
laihenyi/pi-jev-browserTypeScriptAn MCP server that lets Claude hand a whole browser task to Jev: Claude plans and verifies while Jev picks each click, type or select from a numbered table of visible controls in about 100 ms, using browser-use/jev-ultrafast's loop.
Jev Agents & MCP#community#catalog-import#madewithjev
Community
msalvalaggio/jev-reflexPython
msalvalaggio/jev-reflexPythonLets AI agents use Windows applications through MCP or the command line: hand off a bounded task and Jev observes windows, chooses controls, types supplied text and navigates locally, returning a result or a request for help.
Jev Agents & MCP#community#catalog-import#logicrw
Community
jacks3tr/jev-desktopPython
jacks3tr/jev-desktopPythonMCP server that lets Claude Code, Codex and other agents drive a real Chrome browser while JEV makes the small typed decisions — which element, which suggestion, did the click work — with calibrated confidence driving act, verify or ask.
Jev Agents & MCP#community#catalog-import#logicrw
Community
legostin/jev-mcpTypeScript
legostin/jev-mcpTypeScriptAn early-alpha integration that lets TypeSafe Jev choose from concrete browser actions while a host - the Codex sidebar or Browser Use's Python engine - handles generation, recovery, and final verification, without patching the installed Browser Use package.
Jev Integrations & Adapters#community#catalog-import#logicrw
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.