Always-on visual discovery & outfitting AI on every page.
No prompt needed.
Koral builds visual discovery and outfitting into the product pages, images and baskets your shoppers already use. One script tag, or one product ID per request. All it needs is your feed.
- Image search
- See similar
- Matching sets
- Complementary products
- Image search
- See similar
- Shop the look
- Complementary products
- Image search
- See similar
- Shop the look
- Image search
- See similar
- Home improver
Most shoppers never open the chat. Every shopper sees the product page.
A tool the shopper has to open reaches the shoppers who open it. Koral puts discovery and outfitting on the pages every shopper already visits, and serves your assistant from the same engine.
Already there.
Related products, matching sets and complementary products are on the page when the shopper arrives. Nothing to open, nothing to learn.
Whole catalogue.
Sets are generated ahead of time for the products in your feed, in every market and language the feed covers.
In the assistant too.
The same engine serves your shopping assistant over MCP, for the shoppers who do ask.
Vendor data puts on-site chat at about 1 in 100 visits: 0.84% of 4.78 billion e-commerce visits in 2024. Smartsupp, 2025
Shoppers see before they search.
Inspiration arrives as images: feeds, stories, a friend’s living room. The search box still wants words.
- 01
The photo is the query.
A screenshot, a product shot, a room. The shopper already knows what they want. They just can’t type it.
- 02
Words are a bad proxy.
“White teacup” returns thousands of results, not that one. Every extra filter is a chance to leave.
- 03
The images are already yours.
Product, campaign, editorial and user photos: every one of them can be a way into the catalogue.
Every image becomes a way in, and a way to buy.
Koral reads the images you already have, from product and campaign shots to editorial and user photos, and turns them into search, discovery and outfitting on your own catalogue.
Find
Search with a photo, see similar, discover from any content image. The shopper never hits a dead end.
→ More clicks to purchase.
Visual discovery →Complete
Matching sets, complementary products, bundles and finished looks, generated per product and served wherever the basket is built.
→ Bigger baskets and higher order value.
Bundling & outfitting →Run
Rules for what shows where, a dashboard for merchandisers and events in your analytics.
→ Measured directly on add-to-cart and sales.
See the implementation →What it does, measured by the clients.
Conversion in sessions using See similar, 5× the site average of 1.6%.
Higher conversion than the site’s other product carousels.
More revenue per session than the incumbent AI recommender, in a 246K-user A/B test.
More add-to-carts for shoppers who saw a matching set.
More sales from the baskets those sets created.
Of tracked conversions in the pilot were linked to AI use.
Higher conversion when Koral tools were part of the session, compared with the site average.
All numbers from the clients’ own analytics, A/B-split where possible. Method and full figures on request.
The biggest platforms already moved.
Your customers already search with their camera and photos on Google, Amazon and Pinterest, and expect to do the same on your site. Those platforms spent years building it.
Same shift, your catalogue: Koral gives you the capability without the team, the models or the years.
Two families, one engine.
Visual discovery runs from one script tag. Bundling & outfitting is served by the API.
Implementation: this is all it takes.
No model to train, no tagging, no CMS work. The product feed is the only input; the rest is a tag or an endpoint.
<!-- one tag, via GTM or the template --> <script src="https://script.koral.nu/dist/index.js?searchId=YOUR_SEARCH_ID&similar=true"></script>
Widgets are included and take your colours, radius and language. Or render your own components on the same data.
POST https://search.koral.nu/api/v1/search/complementary
{ "productId": "YOUR_PRODUCT_ID", "feedId": "YOUR_FEED_ID" }
200 OK
{ "matches": ["231956529", "231955952"] } Sets are generated ahead of time, so the endpoint just serves them, sub-second. Your team renders them where they belong.
Feed connected
CSV, JSON or your platform feed. Koral indexes every image, in any language and any category.
Tag or endpoint wired
Add the tag through GTM, or call the endpoints from your own components. Your design, your markets.
Live and measured
Pilot on one category, A/B-split, with events in your analytics. Then roll out market by market.
99.9% uptime SLA. The script is served from a global CDN; requests are processed in the EU. Koral keeps nothing from shoppers: photos and data are used for the request and not stored. A data processing agreement (DPA) comes with every contract.
The capability without the build.
Amazon, eBay and Pinterest each built visual discovery in-house over several years. Koral gives you the same capability on your own catalogue, without the team or the timeline.
| Approach | Time to value | What you get |
|---|---|---|
| Build in-house | 8–12 months (Koral’s estimate) | Data science hires, model training and an image pipeline to maintain forever. |
| Legacy recommendation engines | Weeks | Behavioural signals only, blind to what the product actually looks like. Cold start on every new item. |
| Koral | Days | Visual understanding out of the box. One script tag or one endpoint, no model to train, rules and analytics included. |
By some estimates, more than 80% of AI projects fail, twice the rate of IT projects that do not involve AI. A measured pilot on your own catalogue takes that risk off the table before you commit. RAND, 2024
Simple pricing that grows with you.
Any combination of tools: we set the package together, and every tool can be enabled. The four packages below are starting points. Test first on a free demo page; the pilot is 50% off list prices. Prices excl. VAT.
Essentials
On-site product discovery
Search
Image search + discovery
Baskets
Sets, add-ons and outfits
Complete
Search + Baskets
Analytics and rules are included in every package.
Essentials and Search cover one market on web and app. Baskets and Complete scale with product updates per month: one update is counted each time a product from a selected category is ingested from the feed, or receives a major update. Prices shown cover up to 25,000 updates; larger volumes are priced on request. Bundles & Finish the look and AI Stylist are counted per generated outfit, from €0.06 each, lower at volume.
Large catalogues, many markets or the full suite? Enterprise is priced by agreement, with up to 20% off the combined price. Talk to us →
Compare packages and add-ons →Frequently asked questions.
Koral adds visual discovery and outfitting to your store: image search, similar products, shoppable content, matching sets, complementary products and complete looks. Most of it is always on, already on the page when the shopper arrives.
Text search depends on the words a shopper knows. Visual search reads the image itself, including colour, shape, style and pattern, and finds the closest matches in your catalogue.
Any platform. Visual discovery is one script tag, added through GTM or your template. Bundling and outfitting are REST endpoints your team calls with a product ID. Your product feed comes in as CSV, JSON or your platform feed.
Essentials starts at €1,100 per month and Search at €2,400, each for one market on web and app. Baskets, with matching sets, complementary products, bundles and AI Stylist, starts at €1,600 for up to 25,000 product updates per month, and Complete, with both, starts at €3,600. These are starting points: any combination of tools is possible, and we set the package together.
Connect your feed on day one, add the script tag or wire the endpoint on day two or three, and run a measured pilot from the first week.
First we build a free demo page on your own catalogue. The pilot is then one month on your own site at 50% off list prices, usually on one category and A/B-split, with events in your own analytics.
Yes. A REST API with sub-second responses, a 99.9% uptime SLA and full documentation. The script is served from a global CDN; requests are processed in the EU. Matching sets and complementary products are generated ahead of time, so the endpoints just serve them.
Fashion, home & interior, electronics, toys, sports, and more. The engine works across e-commerce verticals.
See Koral on your own catalogue.
Start with a free demo page on your own catalogue, then a one-month pilot at 50% off list prices. All we need is a product feed. Tell us a little about your store and we'll reply by email.
Thanks — we'll be in touch.
We've got your message and will reply by email.
Prefer email? hello@koral.nu