# Partsgraph

> The agent-ready parts data layer. Partsgraph ingests component manufacturers'
> datasheets, parametrics, CAD and compliance documents into one canonical
> Parts Graph — and serves it to AI agents over MCP, agent-readable pages and
> shopping feeds. Monitored for accuracy, measured per agent, deployed as an
> overlay with no website rewrite.

Status: pre-launch. Founding-partner programme onboarding five UK/EU
component and building-product manufacturers (Q4 2026).

## What Partsgraph does

- Ingest: PDF datasheets, PIM/ERP, site crawl, CAD/BIM — eval-gated extraction
  with per-field confidence and source links.
- Canonicalize: typed parametrics per ETIM class, MPN/variant entity
  resolution, cross-reference and alternates graph.
- Serve: hosted MCP endpoint (parametric search, part lookup, alternates,
  compliance, CAD; OAuth for gated price/stock), server-rendered structured
  data + markdown mirrors, GTIN/MPN-keyed AI shopping feeds, ETIM xChange 2.0
  and EU DPP / AAS (IEC 63365 digital nameplate) exports.
- Measure: agent query analytics, share-of-answer, continuous golden-set
  accuracy probes of the major assistants.

## Why it exists

AI assistants and agents now sit inside B2B buying (94% of B2B buyers used AI
in their most recent purchase — Forrester 2026), but AI crawlers execute no
JavaScript and product pages are the least machine-readable page type on the
web. Manufacturers' parametric catalogs are effectively invisible to the
systems doing the shortlisting. Partsgraph is the authorized, accurate,
first-party layer that fixes that — because the alternative is agents
answering from scrapes, mirrors and guesswork.

## Contact

- Early access: see the form on / (we audit your catalog first)
- This site is served agent-readable by design.