One platform, from raw file to trusted record.
Four layers do the work: context that grounds every decision, agents that do the reading and matching, governance on what leaves, and one screen your team actually works in.
Platform overview
Everything the agents know
Where the work happens
What leaves, and on whose say-so
The screen your team works in
Messy in. Structured, checkable, sourced out.
Six agents, one job each.
Each takes a stated input and returns a checkable answer with its evidence. Pick one, or watch the six cycle — the anatomy below always matches the agent shown.
Knowing the rules is the work.
Your catalog runs on rules nobody wrote down: which supplier codes supersede which, what a valid replacement is in your category, which attributes a quote depends on, where a pack quantity hides. Claro does not ask you to re-model your business to fit a schema. It learns those rules from your data and your team's decisions, structures every record around them, and applies them the same way on the first file and the ten-thousandth — so the output looks like your catalog, not a generic one.
Any model can generate an attribute. The work is knowing which value is right for your business — and being able to show why. That judgment, captured once and applied continuously, is what Claro is.
Three lanes, routed by blast radius.
Not by confidence alone.
A high-confidence merge is still a merge. What a change can break decides who signs it off, so the score never buys the last lane.
Auto-apply
no human, reversibleCell-level fixes from a fix-pattern that has already earned the lane. Unit conversions, casing, known normalisations.
Review queue
the defaultEverything not yet trusted enough to apply itself. Grouped by fix-pattern, so a reviewer approves 312 changes as one decision, never 312 rows.
Always human
whatever the scoreMerges, un-merges and reclassification. A confident wrong merge is more expensive than a missing value, so confidence never buys this lane.
Clean once and stop? No. Running every day.
What Monitor sees becomes context for the next run. This is the part most tools do not have.
One canonical record per real product
Every downstream system references it instead of keeping its own conflicting copy.
Trusted product and supplier data is the foundation. Over time, the same layer can power the commercial and operational decisions built on top of it — from procurement and pricing to search, sales and AI.
Integrates with your stack — and fast.
Claro connects to SAP, Microsoft Dynamics, Odoo, Akeneo, Pimcore and more — by API, file exchange or direct write-back. Your existing systems stay exactly as they are. Your backoffice doesn't.
From first call to a working pilot in ~11 business days.
No twelve-month transformation. No "phase 1 of 4." Our solutions team ships a working pilot on your real supplier files in about eleven business days, then you decide whether to roll it out.
Day 1Discovery call
30 minutes with a solutions engineer. We learn your catalog, your existing systems, and the bottleneck worth fixing first.
Day 2Map the workflow
We map a single high-value workflow — supplier onboarding, catalog matching or enrichment — to a concrete agent.
Day 3Sandbox run
Sample supplier files flow through the agent in your sandbox. Outputs are wired to your downstream system.
Day 5Review & tune
Our solutions team reviews the outputs with your catalog team, tuning prompts, fields and confidence thresholds.
Day 10Side-by-side
Claro's output vs. your team's manual review, measured — accuracy, time saved and coverage on your real data.
Day 8–11Pilot live
The working pilot runs on your real supplier files, measured side-by-side against your team's process. You decide, with the numbers in hand, whether to roll it out further.
One layer. Four ways to run it.
The same matching, enrichment and validation layer, delivered however your business already works — on the Claro platform, as an API your application calls, embedded as a feature of your own product, or as an MCP server your agents query directly. We don't change what we do or how we do it; we change where it runs. The best fit is usually a short conversation, not a fixed SKU.
Claro platform
Your team works the catalog in Claro directly — the grid, the review queue, the confidence scores. Nothing to build; onboard a supplier file and start.
API
Call matching, classification, enrichment and validation per record or per batch. Schema-aligned, versioned, deterministic and confidence-scored — the same output shape every time.
Embedded
Claro runs catalog operations inside your platform, for your customers. You ship product matching as a feature of your own product without hiring a matching team to keep it working.
MCP
Your agents query the catalog layer directly: resolve this reference, is this a valid replacement, what is the confidence, where did the value come from. Confidence scores become permissions — 0.97 with three sources can act; 0.62 asks first.
Every mode ships the same schema-aligned, versioned, confidence-scored output — pick the one that fits how your platform already runs, or mix them.
Your catalog data stays yours
Agents work on your data without taking ownership of it. Every action is governed, auditable and controlled by your team.