jev-browser-skill
Readable reference skill for Claude Code and Codex where Jev picks from a numbered table of page controls, with an explainer site.
Overview
Jev Browser Skill is a deliberately small reference implementation, three short files, seven operations and no dependencies, that lets Jev drive a browser from Claude Code or Codex. You name a site and a goal, and Jev clicks, types and selects its way there in a few hundred milliseconds per decision. A demo logs into Hacker News, searches for Jev and opens the first result's comments in nine decisions over eight seconds.
Each step, a snapshot running inside the page returns a numbered table of visible controls with role, label, value and allowed operation, and Jev sees only that table and the page's visible text, never the DOM. It answers a multiple-choice question with a key from the list plus calibrated probabilities, and code executes the pick. An explainer site steps through a recorded trace. The README says it works on ordinary pages and stops at the interesting limits, and points to browser-use/jev-ultrafast for a production-grade version with an inspector, region narrowing, stale-page guards and a text-generating helper. It is MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which numbered control and operation should be used next? | — |
How it works
- Snapshot visible controls into a numbered table
- Jev picks one row and operation
- Code executes the action
- Repeat until the goal is reached
Strengths
- Small enough to read in one sitting
- No dependencies
- Step-by-step explainer with real traces
Caveats
- Reference implementation; limited on complex pages
- Lacks production features found in jev-ultrafast
Getting started
Install as a skill in Claude Code or Codex from the repository; see https://jev-browser.vercel.app for the walkthrough.At a glance
- Type
- Agent skill
- Status
- prototype
- Language
- JavaScript
- Runtime
- agent skill
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Let Jev, TypeSafe's ~100ms decision model, drive your browser. A plug-and-play skill for Claude Code and Codex.
logicrw/awesome-jev-projectsLet Jev, TypeSafe's ~100ms decision model, drive your browser. A plug-and-play skill for Claude Code and Codex.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev see?
A numbered table of visible controls and the page's visible text, not the DOM.
Is it production-ready?
No, it is a reference; the README points to browser-use/jev-ultrafast for that.
How fast is it?
A few hundred milliseconds per decision; the demo took nine decisions in eight seconds.
Related projects
Portuguese agent skill teaching Claude Code, Cursor and Codex to use Jev correctly, bundled with the full raw research behind each claim.
Agent skillproductionSkill that makes Claude Code and Codex split AI features into Jev, LLM or plain-code steps before writing code.
ProjectbetaAgent-plugin browser tool where Jev picks one operation and one row from a numbered page table; a small LLM only writes typed text.
ProjectbetaTypeScriptChrome extension plus MCP server so Claude or Codex can drive your real Chrome, with Jev picking each click in about half a second.
MCP serverbetaTypeScriptJev browser agent that pierces shadow DOM and iframes, adds canvas fallback and full actions, with one Jev request per cycle.
MCP serverbetaPythonMCP server that lets Claude Code or Codex drive real Chrome with Jev picking elements and checking steps, plus safety gates and traces.
MCP serverbetaTypeScript