jev-browser-skill (ChenYCL)
Browser and computer use for coding agents via CLI and MCP, with Jev's calibrated judgments and a code-owned control loop.
Overview
jev-browser by ChenYCL provides browser use and computer use for coding agents, powered by Jev's calibrated judgments with the control loop in code and no vision model or prompt-and-parse. It runs in the user's own browser: the default backend is ego lite, which reuses signed-in sessions and hands control back to the user at logins, CAPTCHAs or consent walls, resuming from the same tab afterwards; Chrome and Safari are also supported. It is available as a CLI and an MCP server, with English and Chinese docs.
Demos recorded by the tool itself include a multi-hop Wikipedia task through a 500-link page (6 steps, $0.0029, 21 s) where goal-aware candidate ordering keeps relevant links in the list; signing in with a secret shown to Jev only as an input placeholder and filling a form with a dropdown (10 steps, $0.0013, 18 s); and choosing the correct Start free trial button among three identical ones from page structure at probability 1.00 (3 steps, $0.0003, 3 s). Each decision is a number you can threshold. It is MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which link, button or field should be acted on next toward the goal? | — |
| Choice | Which dropdown option matches the requested value? | — |
How it works
- Agent runs jev-browser with a goal
- Page candidates are ordered by goal relevance
- Jev picks the action and target
- Code executes; hand to user at logins or CAPTCHAs
- Repeat until done
Strengths
- Reuses the user's signed-in sessions
- Secrets referenced by name, not shown to the model
- Very low cost per task in demos
Caveats
- Default backend ego lite is a specific browser
- Acting in a real logged-in browser needs care
Getting started
Install the CLI or MCP server from the repository and run `jev-browser run` with a goal.At a glance
- Type
- MCP server
- Status
- beta
- Language
- TypeScript
- Runtime
- CLI, MCP server
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Browser use & computer use for coding agents, powered by TypeSafe Jev: calibrated judgments from a System One model, control loop in code. ego lite / Chrome / Safari · CLI + MCP
logicrw/awesome-jev-projectsBrowser use & computer use for coding agents, powered by TypeSafe Jev: calibrated judgments from a System One model, control loop in code. ego lite / Chrome / Safari · CLI + MCP.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Which link, button, field or option to use at each step.
How are passwords handled?
They are typed by code and shown to Jev only as a placeholder name.
Which browsers work?
ego lite by default, plus Chrome and Safari.
Related projects
Local stdio MCP server that executes whole browser and macOS goals with a Jev-driven observe-decide-act loop.
MCP serverbetaTypeScriptMCP server and libraries that drive Chrome over CDP and hand a decision model like Jev a bounded menu of page actions.
ProjectproductionTypeScriptRust MCP server for desktop control via accessibility trees, with an optional decision model (Jev or compatible) for small choices.
MCP serverproductionRust- Jev Browser★ 335
Headless browser automation via MCP server, CLI or library: Jev picks one action per step and scores goal-met and stuck probabilities.
MCP serverbetaTypeScript Rust CLI for desktop computer use via accessibility trees, with optional jev-desktop scripts letting Jev pick the control and action.
Agent skillbetaRustmacOS computer-use package: a background app driver for any agent, plus a loop where Jev picks each UI action for a bounded subtask.
ProjectbetaPython