Upsellr V2: master data at the core, a fully agentic app

Since day one, Upsellr has been the AI agent that automates your product data: it enriches, categorizes, generates and publishes your product pages on your sales channels. More than 800 merchants use it to work thirty times faster than by hand. Today, we are launching Upsellr V2. It brings two fundamental changes: your catalog now rests on master data you truly manage, and the app becomes fully agentic.

What V1 did

The first version of Upsellr automated every step of product content. The agent searched for information on the web and in your documents, categorized your products, generated descriptions, benefits, FAQs and tags for each output format, then published to PrestaShop, Shopify or Amazon. For a catalog, that was already a change of scale.

But product data was still there to serve the product pages: you enriched in order to generate, channel by channel. And it was up to you to chain the steps together, configure each format and relaunch each processing run. With V2, we tackled the problem from the other end.

Master data at the heart of the catalog

In Upsellr V2, every product has its master data: its reference record. Identifiers, brand, technical features, images, manuals and documents all live there, and each value keeps the source it came from. This is what you browse, filter, correct and have completed. Your channels’ product pages are derived from it.

One reference, as many channels as you like. Master data keeps values as they arrive, without bending them to fit any particular channel. Units, formats, categories and fields are adapted to a channel at the moment its product page is generated. Adding a marketplace tomorrow does not mean touching today’s catalog.

Data you manage, not just data you produce. You see what is missing, channel by channel, thanks to each product page’s completeness. Duplicates are caught at import. A correction made in the master data benefits every product page derived from it.

Reliable data. Enrichment favors the manufacturer’s official sources and keeps only what is verified. Nothing is made up: a value that cannot be found stays empty, and you always know where a found value comes from.

A fully agentic app

In V2, you no longer move from screen to screen to chain processing steps together. You write to Marcus, your AI data manager, as you would to a colleague: “Here’s the file from my new supplier, prepare the product pages for my store and for ManoMano.” Marcus knows your catalog, understands the request and organizes the team’s work.

1. 01Marcus, data managerhe imports your files, edits the catalog in bulk, sets up your channels and their templates, and coordinates the other agents.
2. 02Lily, enrichmentshe completes the master data: features, images and documents found on the web and in your sources, each with its source.
3. 03Max, generationhe writes each channel’s product page: title, descriptions, FAQ, SEO tags, category and translations, in your brand voice.
4. 04Clara, analysisshe audits your catalog and your live product pages, and delivers a sourced report, without ever touching your data.

A plan before every action. For a significant request, Marcus builds a workflow, assigns it to Lily and Max, and presents it to you with its estimated cost. Nothing starts until you click “Run”. Nothing irreversible happens without your approval.

Channels set up through conversation. Creating a channel no longer means filling in a long configuration: Marcus asks you the right questions, starts from your industry’s library (families, features, value lists, writing guidelines) and builds the structure with you.

Bulk edits in a single sentence. “Put all Makita drills in the Drilling category”, “Find the products without a main image”: Marcus finds, proposes and applies, across thousands of products at once.

Why it is a breakthrough for merchants

You go from a tool you operate to a team you delegate to. You state the outcome you need; the team of agents takes care of import, enrichment, structuring, writing and publishing.

Months of work become a few days. At Les Mousquetaires, the Équipement de la maison division loaded 139,000 SKUs into Upsellr. Six days later, 127,225 product pages were enriched and published on three websites, with an error rate below 0.5%. Read the customer story.

A catalog ready for every channel, including those that do not exist yet. Clean, sourced master data can be rolled out to a new store, a new marketplace or a new language without starting from scratch.

Visibility in generative engines. Complete, structured product pages are the ones ChatGPT, Perplexity and Gemini understand and cite. At Gédimat, the national website became the first source cited by ChatGPT on the local products tracked. Read the customer story.

Products finally stepping out of the shadows. At Teract (Gamm Vert, Jardiland), 1,749 products existed only in stores. Starting from partial data, the agents rebuilt and published their product pages, without involving the central PIM teams. Read the customer story.

Frequently asked questions