Hey Product Hunt 👋 I’m Matteo, co-founder of Claro.
We created Claro because we kept losing confidence in our AI outputs. @tameesh and I spent years at DeliveryHero and Yelp trying to solve product and supplier data at scale. Tamesh then conducted NLP/LLM research and published in the MIT Review. Every time we plugged LLM into a data workflow, we got pretty-looking answers with no way to know which was correct.
Claro is the AI execution layer for product and supplier data. It takes messy supplier feeds, spreadsheets, and documents, turns them into trusted catalog entities with stable IDs, and keeps them validated and valid over time.
Today we’re opening the first public module: Research Agents. 14 Task-specific agents that run within local tables:
📋 Find your perfect listing. – Define ideal prospects, suppliers, or partners, get an enriched dataset 📍 List the location – Draw an area on the map, create POI with attributes 📄 Capture table from PDF. – Draw tables from reports, choose which one to keep 📑 Convert documents to structured data. – Extract fields in hundreds of documents at once. Scrape data from URLs. – Convert list of links into organized rows 📊 Parse the CSV. – Upload, validate and enrich a spreadsheet 🔀 Merge and map – DIP and merge records from two files and more.
What makes Claro different is what happens to each response before it hits your table:
→ Multimodal consensus – We extract multiple models in parallel and accept only where they agree.
→ LL.M. as Judge Filtering – A lightweight quality gate catches junk output before it reaches you.
→ Classification of sources – First-party and authoritative sources over random blog posts, datasheets over application guides
→ Confidence scores – Each cell is scored at the row, column, and entity level based on source reliability, cross-source consistency, model assurance, and retrieval quality.
→ Full quotes — Click any cell to view the exact passage, URL, or section of the document
→ BM25 + AI hybrid matching – For entity resolution against your existing catalog, with human approval loops for borderline cases
This is not a cheat wrapper. It’s the first validation infrastructure with a spreadsheet you can use today.
Three things I would like to do today:
Try it out – 200 free credits, no card.
Break it up — Throw in your ugly CSV or weirdest PDF, tell me what went wrong (and what the trust score said)
Let me know which agent you will use first. – We are deciding what to do next.
I’m pitching live at The Pitch Berlin today, so I’m bouncing between the stage and the comments, but I’ll respond to everyone before EOD.
DM me if you run a marketplace, distributor, or multi-supplier catalog and the core platform sounds relevant — we’d love to talk.
Thanks for taking a look 🙏 – Matteo