Why not just ask a big LLM what to do from screenshots?

That loop is slow and expensive; here Jev decides from DOM state in one structured call and a small model writes text only on text-entry steps, which is what keeps runs fast and cheap.

What setup pain should I expect?

You need Chrome with remote debugging on a fresh user-data dir, a WebSocket URL that expires every session, plus both a TypeSafe key and an OpenRouter key in .env.

Will it book a flight with my card?

No. The agent stops at the results page and the README warns never to let it proceed to payment or booking without explicit human confirmation.

Any gotchas on Windows or nonstandard installs?

Yes: set UTF-8 encoding for text reads, use the required Chrome flags and inspect toggle, and refresh the debugging URL each session.

What does a flight search cost and how long does it take?

About 7 seconds per search and roughly $0.0039 per run, per the README. Jev's output tokens are free and only the small text model is billed for generating strings.

What stack does the flight agent run on?

Jev as the decision layer, Playwright for browser automation, browser-harness for the CDP connection to Chrome, and Python 3.12.

How do I start the agent after configuring the keys?

Launch Chrome with remote debugging on port 9222, read the WebSocket URL from the version endpoint, export it as BU_CDP_WS, run uv run jev, then open 127.0.0.1:8766 and click Start demo.

CommunityJev Agents & MCP

Jev Flight Agent

Browser-use agent that completes flight searches in real Chrome: Jev picks each next action from an indexed DOM table while a small model handles only text entry.

See highlights

Open github.com

Stars: 0
Language: Python
License: MIT
Published by: Community
Author: kallurayaankit
Last verified: Oct 2026

kallurayaankit/jev-flight-agent

User repository on GitHub

View on GitHub

Browser-use agent that finds flights with Jev (TypeSafe AI). First project with TypeSafe — Jev picks, code acts.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A browser-use agent that finds flights using Jev, TypeSafe AI's System One decision model. Built as my first project with Jev. Give it a natural language task like "Find one-way flights from Zurich to London on September 20, 2026" and it drives a real Chrome browser to complete the search. Jev picks the next action at each step. A small language model handles only text generation (like typing "Zurich" into a field).

Quoted from the project's README. Read the full README.

Tagscommunity github

Works forJev for Tool calling

Used inJev for Software & DevTools

Languages: Python

Highlights

Reads interactive elements as structured data instead of screenshots, so Jev returns element and operation decisions with confidence.
Keeps text generation on a tiny model that fills fields only when Jev decides text is needed; Jev never touches the browser.
Automates over CDP with Playwright and browser-harness, starting from a served page where you click Start demo.
Stops at the results page by design and never selects or books flights without explicit human confirmation.

Watch out

Demo agent that drives real browsers; keep it away from payments and bookings. 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

agent-guard

Claude Code PreToolUse hook that screens Bash, file edits, and tool calls with Jev risk, policy, and allow-ask-deny judgments before execution.