Bucepha AI Chrome extension See what the auction photos hide. Before your dealership buys.

AI that instantly finds flaws across auction vehicle photos, helping car dealers spot problems, estimate recon, and make smarter bids.

It is a Chrome extension. Open an auction listing, and the report opens beside it: the damage marked on the photograph it was found in, the repair costed, the VIN decoded, all without leaving the page.

Add to ChromeStart free trialFor dealership buyers

01How it works

One press, beside the listing you are already reading.

AI that instantly finds flaws across auction vehicle photos, helping car dealers spot problems, estimate recon, and make smarter bids.

BUCEPHA
Analyzed12 photos
Auction vehicle with the defects Bucepha found numbered on the photo
Auction price
$24,100
Est. repairs
$2K–$4K
Est. margin
$7,100Profitable

The panel as it appears beside a listing. Shown for illustration; nothing here is live.

  • 01

    It finds the car on the page

    Open an auction listing and the panel appears beside it, already knowing the vehicle, its photographs and what a report would cost. Nothing is read and nothing is charged until you ask.

  • 02

    The image model reads the photographs

    A photograph costs 5 credits, charged as that photograph is read. A photograph we cannot read is not charged for, and the panel says how many went unread rather than quietly rounding.

  • 03

    Each defect is marked where it was found

    A numbered marker sits on the photograph the defect was found in, so you see the dent rather than read about it. The marker names the panel, the severity and the repair behind it.

  • 04

    Repairs are priced, and the price is yours to change

    Every operation is booked in tenths of an hour at your own posted labour rates. Disagree with a figure and type over it: the total, the margin and the breakdown follow, and your correction is kept.

  • 05

    The money, in a column you can check

    The bid from the page, the repairs at their likely figure, and the retail price: sourced from the web when the web says, and yours when you press the pencil on it and type. The margin is that figure less the bid less the repairs, coloured green when the deal clears even at the high repair figure, red when it is under water even at the low one, and yellow when the range decides it.

02What it is, in words

A thin client on the listing, and everything that thinks stays on our servers.

The extension is a small Chrome and Edge extension that reads the auction page you already have open. It does not carry a model, a prompt, a price table or a key. What it sends is the listing as the page shows it: the title, the VIN if the page prints one, the bid, and the addresses of the photographs. What it gets back is a report. The reading, the pricing and the scoring happen on our servers, which is what keeps the method ours and the extension light.

Finding the car
Each auction site has its own reader, which knows where that site keeps the title, the VIN and the gallery. A site without one falls back to a general reader that looks for a vehicle on the page. Either way the extension asks before it reads a site for the first time, and it asks again before a report is generated, so nothing is read and nothing is charged until you press the button.
Reading the photographs
Our servers fetch each photograph from the auction site, check the address first, and send it to the image model one at a time. A photograph costs 5 credits, taken as it is read. One we cannot fetch or cannot read costs nothing, and the report says how many went unread and lowers its own confidence to match.
The same listing twice
Auction pages redraw constantly. A listing you have already had read is returned from the report we hold, and a gallery that has grown is charged only for the photographs that are new. A car that comes back to auction under a new lot number is recognised by its VIN, where the page publishes one, and its photographs are not bought again.
What the extension keeps
A session token for your account, the panel's size and position, and your labour rates by ZIP so the next report is priced your way. Photographs are never stored in the browser. The report itself lives in your account and is listed under History, so the same car opens the same report on any machine you are signed in on.

The intelligence behind the panel

Three things read a photograph. Our computer vision model finds where the damage is on the panel. A pretrained language model describes what it is and how bad. Our own records of what the repair actually cost, corrected by dealerships in the panel, decide what it is worth. The first is published on Hugging Face and the third stays private, and the page below says which part is which.