Does this replace the planner model entirely?

It does not replace the large model. Jev picks operation and target from an indexed element table in one request, while code owns observation, stale checks, execution, verification, and exit conditions, with a small text model used when wording is needed.

What stops it from clicking something dangerous or stale?

Stale refs, obscured or unavailable elements, cross-frame misses, and payment, deletion, or unsubscribe targets are refused with zero dispatches. Exit uses explicit until-checks in code rather than model self-judgment.

How do I install it without breaking my current setup?

Install via the skills CLI, the Claude Code marketplace, or a clone with an install script that can also take over the official browser entry. An offline end-to-end test injects DOM on a blank page with no key or network.

Are the reported speed wins real across all tasks?

Yes for linear flows on the measured tasks, with faster per-decision times and cheaper observation and dispatch than the old path. Native selects favored the baseline in the paired rounds, and form filling showed no difference.

How do I install ego-decision-layer?

One line: npx skills add jiangkoumo/ego-decision-layer, or add the Claude Code marketplace and install the plugin; a clone can run ./install.sh, which links the CLI and by default takes over the official ego-browser entry.

What did the 0.4.0 rename change?

ego-jev became ego-decision-layer in 0.4.0 so the name no longer binds one vendor's backend; the migration script is idempotent, the old CLI remains a compatibility shim, and the EGO_JEV_* environment variables are unchanged.

How does it record which model actually answered?

Each request sends a floating alias, so only the response's model field identifies the served version; renderJevSummary prints staged timings and that serving model every step for audit and reruns.

CommunityJev Demos & Experiments

Ego Jev — jiangkoumo

Pluggable browser-automation decision layer that replaces per-step large-model calls with one Jev request for operation plus target, leaving observation and guards in code.

See highlights

Open github.com

Stars: 5
Language: JavaScript
License: MIT
Published by: Community
Author: jiangkoumo
Last verified: Oct 2026

jiangkoumo/ego-jev

User repository on GitHub

View on GitHub

Pluggable decision layer for the ego lite browser: one System One (Jev) call per step replaces the per-step LLM turn, and the backend can be swapped for a local OpenAI-compatible model. Fail-closed execution guards. The measured one — raw bench data, 16 suites, changelog with corrections.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

This is one of several Jev listings for the same project. The canonical entry is Ego Jev — ZephyrDeng.

Tagscommunity awesomejev catalog-import

Works forJev for Classification

Used inJev for Software & DevTools

Languages: JavaScript

Highlights

Answers operation and target element together from one indexed table, including container regions, frames, and shadow roots.
Keeps observation, bare-dispatch execution, fail-closed guards, and exit checks in code rather than prompts.
Reports staged timings and the serving model version with every step for audit and reruns.
Supports credential file chains, overridable backends with a closable fallback, and offline DOM self-tests.

Watch out

Benchmarks are small-sample A/B and paired-round comparisons with raw data in the repo; rerun the scripts for your own tasks. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

laya-vs-jev-arena

Side-by-side game harness where Laya running locally and Jev over API answer the same typed questions for snake, kombat, runner, flappy, tetris, and human-versus-model matches.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

PromptEngineer48/laya-vs-jev-arenaJavaScript

jev-tetris

Live versus Tetris where Jev faces Claude Haiku 4.5, Gemini 3.8 Flash, or local open-weight Laya on the same seeded piece sequence; cleared lines arrive as garbage rows and the first to top out loses.

Jev Demos & Experiments#community#awesomejev#catalog-import

Community

trungdq88/jev-tetrisJavaScript

live-jev

Browser top-down driving sim that sends sensor JSON to Jev every 200ms with four typed driving questions in one call, then applies confidence gating and reflex rules to steer.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

vinilana/live-jevJavaScript

sift

Chrome extension that labels each X post with Substance, Humor, Chit-chat, Promo or Junk plus AI-written and off-topic flags, collapsing hidden posts to one line.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

bohutang/siftJavaScript

xtags

Chrome extension that labels each X post with intent plus rage-bait, synthetic, and undisclosed-ad signals from Jev, with threshold-only rendering and local caching.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

manifoldor/xtagsJavaScript

jev-weekend-shopping-chrome

Chrome extension that judges Taobao and JD listings against double-weekend-off clues and stamps a PASS mark over product images it guesses come from non-weekend-off shops.

Jev Demos & Experiments#community#awesomejev#catalog-import

Community

littlewindy123/jev-weekend-shopping-chromeJavaScript

From the Jev guides

Original write-ups on what this entry is for.

Per-Tick Decisions in Games

Sep 28, 2026

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

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai