Skip to content

jev-ego

TypeScript browser agent for ego lite where one Jev request picks the operation and indexed target element each step.

CLIbetaCustom license

Overview

jev-ego is a TypeScript browser agent for the ego lite browser that reuses Jev Ultrafast's indexed action space without Chrome, Playwright or Browser Harness. Each observation produces a numbered table of actionable elements, and one TypeSafe request picks an operation (CLICK, TYPE_TEXT, SELECT, SCROLL_UP, SCROLL_DOWN, WAIT, DONE or BLOCKED) along with speculative target questions for click, type and select, so two decisions take one round trip.

Only supported operations and targets are offered, and model output never becomes selectors or coordinates; a small LLM supplies typed text when needed. The whole loop runs inside one ego-browser Node process. Agents can drive the page themselves through observe and act commands or ask Jev for ranked suggestions through suggest and step. Uploads, tabs, dialogs, canvas, shadow DOM and iframes are left to the ego-browser skill.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich operation should be performed next?CLICK · TYPE_TEXT · SELECT · SCROLL_UP · SCROLL_DOWN · WAIT · DONE · BLOCKED
ChoiceWhich indexed element is the target for the chosen operation?—

How it works

  1. Observe the page and build a numbered element table
  2. One Jev request asks operation plus speculative targets
  3. Execute the target matching the chosen operation via CDP
  4. Use a small LLM for text when typing; repeat until DONE or BLOCKED

Strengths

  • Two decisions in one round trip
  • No Chrome or Playwright dependency
  • Model output never becomes raw selectors

Caveats

  • Tied to the ego lite browser
  • Complex page features handled elsewhere
  • Custom licence

At a glance

Type
CLI
Status
beta
Language
TypeScript
Runtime
ego-browser Node process, CLI
Has code
yes
License
Custom license

Use cases and integrations

How curated lists describe it

  • Browser agent on [ego lite](https://lite.ego.app/): one TypeSafe request picks operation + indexed element; agent-facing observe/act/suggest/step CLI

    AnotiaWang/awesome-decision-models
  • A browser Agent for ego lite that numbers actionable elements for Jev to choose the next step.

    logicrw/awesome-jev-projects
  • TypeScript browser agent for ego lite: Jev Ultrafast indexed actions, TypeSafe Jev decisions, persistent observe/act CLI. No Chrome or Playwright.

    hellogumbo/awesome-jev

Frequently asked questions

What does Jev decide?

The next browser operation and which numbered element it targets.

Does it need Playwright?

No, it runs inside ego lite using CDP.

Who generates typed text?

A small LLM; Jev only chooses operations and targets.

Related projects

  • RAM-efficient Rust coding-agent harness, listed as using Jev decisions for memory pruning, browser navigation and voice routing.

    ServiceproductionRust
  • MCP server and libraries that drive Chrome over CDP and hand a decision model like Jev a bounded menu of page actions.

    ProjectproductionTypeScript
  • Headless browser automation via MCP server, CLI or library: Jev picks one action per step and scores goal-met and stuck probabilities.

    MCP serverbetaTypeScript
  • Local browser agent that picks each step's operation and element with a Jev-shaped model such as a fine-tuned Laya, no cloud needed.

    MCP serverbetaPython
  • Claude Desktop MCP extension that drives Brave or Chrome with Jev choosing on-page actions and a guard freezing irreversible clicks.

    MCP serverbetaTypeScript
  • Browser-agent benchmark with five timed events from one prompt, plus a ~200-line DevTools agent where Jev picks each action.

    ProjectbetaJavaScript