The canonical data protocol for industrial parts

Every manufacturer names it differently.
We give it one name AI can trust.

808,000+ industrial products across 33+ manufacturers — contactors, drives, breakers, and more — normalized into a single schema. Every fact traces back to the manufacturer's own product page, so agents built on top of it query structured truth instead of guessing at scraped text.

ABB"Rated Operational Current – In-Line Connection (Ie)"
Siemens"operational current / at AC-3 / rated value"
Canonical key
rated_current_ac3_a
sourced to the manufacturer's own product page
808K+Products indexed
5.5M+Verified spec facts
2.8M+Documents archived
33+Core manufacturers
93%Manufacturer-verified
151Product categories

The Problem

Industrial catalogs don't agree on a vocabulary. AI agents can't parse around that.

A person browsing a manufacturer's site can figure out that "Ie" and "operational current at AC-3" mean the same thing. An agent calling a tool can't — it needs one schema it can write a condition against, like rated_current_ac3_a >= 40. Every catalog is its own inconsistent, often JavaScript-rendered site, with its own vocabulary, its own units, its own way of hiding a discontinued part. That inconsistency is a minor annoyance for a human. For an AI agent sourcing a part number, it's the difference between a real answer and a hallucinated one.

The Method

How the translation actually happens

No per-product LLM guessing. The vocabulary translation is designed once per source and reviewed by a person — then applied the same way every time.

01

Collect from the source

Data is collected directly from each manufacturer's own site or catalog. The raw value is kept exactly as published — nothing is normalized away at this stage.

02

Translate by reviewed rule, not guesswork

A mapping dictionary — drafted with LLM assistance, reviewed line-by-line by a person — translates each source's labels into canonical keys. It's applied deterministically, so results are reproducible and auditable, not re-guessed per product.

03

Serve with proof attached

Every fact carries its source key, a confidence score, and a link back to the manufacturer's page. Versions are kept, so a spec change or discontinuation is a verifiable diff, not a guess.

Three Audiences

Built for three audiences

Same underlying data, three different ways in.

MCP / API

For Developers & AI Teams

Give your procurement copilot, BOM tool, or engineering assistant a source of truth it can cite — cross-brand equivalents, structured specs, and datasheets, queryable through MCP.

See the tools →
Substitution

For Distributors

Cross-reference a real substitute in seconds when a line is discontinued or out of stock, fill catalog gaps, and get ahead of spec changes before a customer finds them first.

See what's available →
Competitive intel

For Manufacturers

See exactly how your parts get cross-referenced against competitors, make sure your own specs are the verified source, and get ahead of EOL communication.

See what's available →

Trust

Trust is the product, not a footnote.

93% of products in the database trace their listed specs to the manufacturer's own page — not a distributor's re-listing. The rest are shown, labeled, and queued for direct verification, never hidden.

Read the methodology →

Have a use case in mind?

Tell us what you're building — takes two minutes.