JevHub · jevhub.sh

The captured page is only chrome: nav labels, a $JEVHUB contract address, a chart link and a panel that says booting. No architecture is documented there. The entry copy does record the review shape (verdict, security risk, severity, quality, slop, an honest message, a policy gate), a zero-dependency jh CLI for Node 20+ and an MCP server, but never which model answered.

1,000 files and 25 MB per static site, 1,000 Jev calls per site per day, and 100 calls per visitor. The per-visitor cap is the one to watch: a popular public site burns through it long before the daily site cap, and the page does not document what happens at the ceiling (throttle, hard stop, or upgrade prompt).

Nothing on the captured page describes the certificate. No signing key, no format, no verification command, no statement about whether it covers the commit hash or the branch name. The copy asserts a signed certificate per branch move, so treat it as a product claim until you can inspect one; the page gives you no way to.

The page does not say. It presents jevhub.sh as a hosted git service with signup implied, and the footer links OpenJEV, a skill.md, llms.txt and a node API. There is no documented self-host path, no export or migration command, and no statement about who holds the repos or what happens to them if the project stalls.

The page shows a $JEVHUB contract on Robinhood Chain, CA 0xB22D3A5142B08FBe987a9f52EB6a2Ec5d3ca59D2, with a copy button and a link to a chart, and pairs it with the tagline git for agents, reviews by OpenJEV. Nothing states what the token does, whether it pays for Jev calls, or whether any part of the service requires holding it.

Labels only. The captured page shows those five nav entries, a Review a diff action, a Publish button, and a Docs link, over a panel whose entire body reads booting. No pricing, no repo or site counts, and no description of what Muse, Agents, or Events are appears anywhere in the capture, so treat the sections as unverified product surface.

The footer points at docs, skill.md, llms.txt, and a node API, which is the shape of a host meant to be read by agents rather than clicked through. None of those documents is captured here, so the request format, the authentication, and whether the review path can be called directly all stay unverified. The same applies to the signing story behind certificates.