When is this better than a plain script or an LLM agent?

Use it when you want verifiable navigation with citations and gates around submit, pay, delete, and send. For single-file form filling you already script reliably, a script remains simpler and cheaper.

How is this different from other Jev-based navigators?

Both choose with Jev, but this adds cited answers, validated data, scoped credentials, and an authorization gate. Navigation-only comparisons do not settle which fits every workflow.

What happens if the judging call is slow or fails?

The keyword pass is instant and Jev re-ranks behind it, so rows settle into judged order. If TypeSafe is slow or down, keyword order stands and the footer says so.

How do I keep passwords and payments safe during runs?

Type secrets with declared origins or vault logins so models see names only, use profiles for signed-in sessions, and require the authorize flag before any submit, payment, deletion, or send.

What runtime and license does it use?

Python 3.13 or 3.14, run through uv (uvx fastbrowse) or installed with uv tool install fastbrowse. It is MIT licensed and marked beta, and a Node program can use it without uv or Python.

How does the authorization gate work?

Without --authorize a run stops at needs_confirmation before any submit, pay, delete, or send; the flag allows those. Code, not the model, controls credentials, authorization, and spending.

What does a run output?

Steps go to stderr; the status, cost, step count, and answer go to stdout. --json prints every step, the verbatim page quotes behind the answer, and cost by component, and --max-steps/--max-dollars can bound the run.

CommunityJev Demos & Experiments

Fastbrowse

Browser agent where Jev picks each action from indexed page candidates while an LLM plans and reads, returning cited answers or schema-checked data with credential controls.

See highlights

Open github.com

Stars: 114
Language: Python
License: MIT
Published by: Community
Author: agent-labs-dev
Last verified: Oct 2026

agent-labs-dev/fastbrowse

Organization repository on GitHub

View on GitHub

A fast browser agent: Jev picks each action from what is on the page, an LLM reads and plans, and every claim in an answer cites a quote from the page.

Forks: 14
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Run browser tasks locally or in a cloud browser. Jev chooses from indexed page controls; fastbrowse returns cited answers or schema-validated data. Code controls credentials, authorization and spending. fastbrowse indexes the page into candidates and has Jev, a choice model, pick one, so it cannot click something that was never on the page. Factual claims in a complete answer cite verbatim quotes from the page.

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

Tagscommunity jevbest catalog-import

Works forJev for Classification Jev for Routing Jev for Tool calling

Used inJev for Software & DevTools

Languages: Python

Highlights

Has Jev choose one indexed control per step so it cannot click something never on the page.
Cites verbatim page quotes behind answers and validates structured data before your app acts.
Keeps credentials out of model view with origin-scoped secrets, profiles, and vault logins.
Stops at needs_confirmation unless you pass authorize, with step, spend, and quote detail on demand.

Watch out

Needs API keys and a browser backend, and cloud browsers add their own charges beyond model spend. 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

jev-ra

Browser-use layer for CLI coding agents: Jev picks the operation and target element in one round trip per step, exposed over MCP with a CLI, a Python API and an HTTP server.