Jev Ultrafast
browser-use/jev-ultrafastReviewed by Radar
What it does
Jev Ultrafast is a small demo and Python library from Browser Use. You give it a goal in plain words, and it clicks, types and scrolls in your Chrome browser to reach it, choosing each step with TypeSafe's Jev model and using a small text model only when it has to type. It is for developers testing browser automation. You need Python, the uv tool, Chrome with remote debugging allowed, and two paid API keys (TypeSafe and an OpenRouter-style text model). The README calls it an MVP with known limits.
How easy it is to install
C · install 1 or 2 tools · Windows
There is no installer: you clone the code, install Python 3.12+ and uv, add two API keys to a .env file, run it from a terminal and allow remote debugging in Chrome.
Numbers on GitHub
- Stars
- 23,991on Oct 31, 2026
- Downloads
- —all releases
- Growth
- +292stars in 5 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Python, Other tools
- To build from source
- Python
- Disk space
- 100 MB–300 MB · estimate
- Accounts or keys
- TypeSafe API (Jev model), TYPESAFE_API_KEY, OpenAI-compatible text model, OpenRouter in the example (TEXT_MODEL_API_KEY)
- Cost
- Free, API costs extra · The code is free (MIT). The README says live runs "make paid API calls" to TypeSafe and to the text model provider; no prices are stated.
- License
- MIT (Permissive)
- Can replace
- Browserbase, Manus
- Last update
- Oct 31, 2026
Risk signals (4)
- AttentionVery new repository with many stars
23695 stars in 40 days
automatic check
- Attentionagent uses your real chrome profile
README.md: "Allow remote debugging in Chrome when prompted." and "Owned tabs share the existing Chrome profile." The agent works inside your normal browser, where you may be signed in to sites.
AI review
- Infopage text sent to external models
README.md: each step sends the page's element table in "one TypeSafe request", and the agent will "Send visible text" of the page; typed text comes from a separate model at TEXT_MODEL_BASE_URL=https://openrouter.ai/api/v1 (.env.example).
AI review
- Infoearly mvp short activity
README.md: "Shadow roots, frames, canvas, uploads, pop-up tabs, nested scrolling, and arbitrary keyboard widgets remain outside this MVP." No GitHub releases; the repository was created on 2026-09-16 and the last commit was on 2026-09-18.
AI review
Evidence is shown in its original language.
What Radar could not check
- Whether it works on Windows; the README does not mention any operating system
- Real API cost per task with TypeSafe and the text model
- How reliable it is beyond the few demo tasks measured by the authors
- What the agent might click or submit on sites where you are signed in