The Vindex Atlas

One identity for every car.
Counted where the VIN runs out.

The Vindex Atlas gives each make, model line, year, trim and build a permanent key, issued from public evidence. Where a VIN can’t say which trim a car is, Vindex lists every trim observed on it, with how often each was seen. It never picks.

2025 Mazda CX-5 · check digit 1 · verified

1,646,113Keys issued in the Atlas

7,346,897Trim sightings behind the shares

26,835Patterns with more than one trim

One VIN, resolved4T1BF1FK9GU572575
Check digit
Position 9 reads 9; the other sixteen compute 9.Valid
Pattern
4T1BF1FKGUPositions 1 to 8, 10 and 11: what every car of this build shares. The check and the serial drop out.
Model year
Position 10 reads G: 1986 or 2016. The pattern binds 2016.

Down the Vindex ladder

  1. MarquemakeToyotavxq_w68h5mgRuled at curation
  2. LineagemodelCamryvxl_nwjz915llParent edge
  3. Vintageyear2016vxv_0nf9sdwcrBound to the pattern
  4. GuiseseriesASV50L / GSV50L / AVV50Lvxg_0jf96rbcdBound to the pattern
  5. CutbuildWhere the VIN stopsKnown: sedan, gasoline, 4×2. The trim is not in the VIN.

Trims observed on this pattern

  • SE0.64n 38
  • LE0.15n 8
  • XLE0.15n 8
  • XSE0.03n 1
  • Edition0.03n 1

56 observed vehicles. Each figure is the share of observed cars that stated the trim, not the share built. Vindex lists every one; it doesn’t pick.

The VIN runs out before the car does.

Positions 4 to 8 are the maker’s to assign, and many makers stamp one pattern on several trims.

So a decoder that returns one trim is guessing. Vindex counts what was observed on each pattern and posts every candidate with its share. Every share on this page is of observed cars, never of production.

Trims stated per VIN pattern3,724,243 vehicles

60,765VIN patterns observed

26,835carry more than one trim

7,346,897trim sightings behind the shares

  1. 33,9301 trim
  2. 14,0862 trims
  3. 6,2243 trims
  4. 2,9624 trims
  5. 1,4855 trims
  6. 8176 trims
  7. 4057 trims
  8. 2718 trims
  9. 1649 trims
  10. 11110 trims
  11. 31011 to 44 trims

A pattern is the VIN with its check digit and serial set aside. The tail runs to one pattern stated as 44 different trims.

Colours, counted the same way3,419,533 vehicles
  • White27.7%
  • Black21.1%
  • Gray20.7%
  • Blue8.9%
  • Silver8.5%
  • Red7.0%
  • and 8 more families

Paint names stay as the maker states them; the family is counted, never guessed.

Issued on evidence, or not issued.

The Atlas is built clean, from public evidence, with no licensed database anywhere in its ancestry.

Every identity is issued once and keeps its key for life: a rename is a relabel, a correction is a new fact, and nothing is deleted. Where the evidence stops, the gap is recorded and counted, never patched over.

The Vindex ladder, issued1,646,113 keys
  1. Marquemake12,240Toyotavxq_w68h5mg
  2. Lineagemodel31,608Camryvxl_nwjz915ll
  3. Vintageyear477,1562016vxv_0nf9sdwcr
  4. Guisetrim or series501,488ASV50L / GSV50L / AVV50Lvxg_0jf96rbcd
  5. Cutbuild623,621This Camry stops above the cut: its trim is not in the VIN

1,602,265 parent edges hold the levels together

One key, taken apart2016 Toyota Camry

vxv_0nf9sdwcr

vxv_
The kind: vintage, the model year of a model line.
0nf9sdwcr
An opaque code for node 345,758, written in digits and consonants only, so a key never spells a word. It carries its own check, like position 9 of a VIN.
  • One character changedvxv_0nf8sdwcrRejected: the check fails
  • Two characters swappedvxv_0fn9sdwcrRejected: it does not decode
  • The wrong kindvxl_0nf9sdwcrRejected: the kind does not match

All 252 one-character typos of this key are rejected, and all 8 swaps of neighbouring characters: a typo fails closed instead of naming another car.

785,693Candidate sets recorded, never guessed

1,022,525Readable addresses, one per identity

12,925Maker codes in the registry

1,515Lines ruled by hand at curation

One ladder instead of 28 maps.

8 catalogs joined pair by pair need 28 hand-kept maps. Through one ladder, each links once.

A catalog’s identifier stays its own: Vindex never rewrites it. It posts a link from that identifier to one rung of the ladder, with a confidence written as a number, and every pair of catalogs meets there.

8 catalogs

28maps, each kept by hand, each drifting on its own

8links, one per catalog, every pair meeting on the same ladder

One link, up closevxv_0nf9sdwcr

The 2016 Toyota Camry, linked to four records in the federal fuel economy data. The records stay the government’s; Vindex posts the link and its confidence.

  • epa-fueleconomy:vehicle:367951.00
  • epa-fueleconomy:vehicle:367960.85
  • epa-fueleconomy:vehicle:367970.85
  • epa-fueleconomy:vehicle:367981.00

89,975federal links posted, each with its confidence

91.62%of 50,242 fuel economy records joined to the ladder

Read the way its market reads it.

The first character names the region. Each market decides whether position 9 is a rule.

A European VIN can fail a US-style check digit and still be a real car; a Japanese car’s domestic identity is not a VIN at all. Vindex validates each VIN by its market’s own rules, and posts the market as a fact beside the key, never inside it.

The first character, the regimeISO 3780

North America

Europe

Asia

Oceania

South America

Africa

JM3KFBXY1S0597748The CX-5 on every Vin page starts here.

J

Japan (ISO 3780 J block)

Domestic identity is not the VIN

MLIT/NALTEC regime: domestic identity is katashiki + frame number

  • Check digit enforcedUnited States, Canada, Mexico, China
  • Check digit not enforcedEuropean Economic Area, Germany, France, Italy, Netherlands, Finland, Sweden, Spain, Great Britain, Korea, Australia
  • Domestic identity is not the VINJapan

Free to read.

The ladder and its counts read free, with no account.

  • FreeThe ladder, the keys, the counts$0
  • Per resolutionOne VIN resolved, its candidates scoredPosts live
  • Pair tableOne catalog pair’s mapping, licensedPosts live
  • Full mapThe whole crosswalk, licensedPriced on paper

Not yet open · each cross-mapped provider's license must name crosswalk redistribution before mapped IDs transact

Provider-mapped IDs go on sale as each provider’s license clears; the ladder and its counts stay free.

Where it’s sold

When a per-resolution price posts, every paid resolution returns a machine-readable offer (posted amount, x402 settlement, expiry), so an agent settles per call on its own account, reading this same page.

The same identities, at work.

Every door that reads a car reads it on this ladder.

The other addresses read the same seventeen characters.

For the agent you send.

The agent reads the same record you do.

Same two routes, three representations: the numbers never differ between them.

Three representationsvindex.vin
1curl https://vindex.vin/ -H 'accept: text/markdown'
2 # the markdown twin of this page
3curl https://vindex.vin/{VIN} -H 'accept: application/json'
4 # the record as typed evidence objects
5curl https://vindex.vin/{VIN} -H 'accept: text/markdown'
6 # the record's markdown twin
On this host
  • Motionsb2aThe buyer is an agent: no human leaf; discover, bind, pay, worst-case cost posted before the call
  • One MCP serverapis.vin/mcp serves the whole platform: every door’s tools in one server, each result carrying its door’s URL.
  • AttributionX-Client-Type: html | md | json | mcp names which reader arrived; it never changes the record.