Scored reads, ranked alternates

VIN OCR API

Pull a VIN out of a photo, with a score.

  • The 17-character VIN, plus confidence
  • Ranked alternate candidates, each scored
  • Bounding box around the detected text
  • Public image URL or a base64 string
Test APIDocumentation
VIN OCR API

Our benefits

For the moment a VIN gets typed in wrong.

A read you can gate on

Every extraction returns a confidence score and ranked alternates, so a low read can trigger a retake instead of a wrong VIN entering your system.

Integrate in minutes

One POST with an image. Copy the request from the panel below, add your key, and the first extraction runs in minutes.

People who know the model

A free sandbox, a documented response, and support that will run your own photos before you design the capture step around them.

Capture however you like

Send a public image URL, or a base64 string straight from the device, so a photo taken in the bay never has to be hosted first.

For developers

A photo in, 17 characters out.

The panel on the right is a real request with an image URL and the response it returns. Scroll it to see the confidence score and the alternate candidates, then copy the request into your terminal and point it at a photo of your own.

Alternates come ranked
candidates carries the other readings with their own scores, which is usually where a misread O or 0 shows up.
404 means two things
Either no valid VIN was found in the image, or the API key was not found or is disabled. Check both.
URL or base64
image_url takes a public URL or a base64-encoded image string in the JSON body.
Request
curl -X POST "https://api.carsxe.com/v1/vinocr?key=YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"image_url":"https://example.com/vin-plate.png"}'
Response
{
  "vin": "JHLRD77874C026456",
  "confidence": 0.94,
  "box": { "xmin": 112, "ymin": 340, "xmax": 498, "ymax": 389 },
  "candidates": [
    { "vin": "JHLRD77874C026456", "confidence": 0.94 },
    { "vin": "JHLRD77874CO26456", "confidence": 0.52 }
  ]
}

For AI agents

Built into the AI tools you already use

Point your coding agent at the CarsXE MCP server and it can call the API on its own, no client library, no glue code. The same tools are available to chat agents over plain HTTP.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Grok Bot
  • Codex
  • Gemini
  • Hermes Agent
  • OpenClaw
  • opencode
  • & more
Explore integrations

One command. Register the CarsXE MCP server with your editor.

Claude Code
claude mcp add --transport http carsxe https://mcp.carsxe.com/mcp \
  --header "X-API-Key: YOUR_API_KEY"
View the docs

Then just ask. The agent picks the right CarsXE tool and returns structured vehicle data.

Prompt
Extract the VIN from this windshield photo

Agent onboarding. Are you an AI agent? Fetch this skill to get an API key and start pulling vehicle data.

SKILL.md
curl -s https://carsxe.com/agent-onboarding/SKILL.md
View the skill

Fair billing

No data, no charge

A lookup we cannot resolve comes back as a 404 and costs you nothing. You are billed only when the API returns data.

Data found200 OK
Data found
{ "vin": "JHLRD77874C026456", "confidence": 0.94 }

Billed. One lookup.

No match404 Not Found
No match
{
  "success": false,
  "message": "No valid VIN candidates found in the image"
}

No charge.

Priced per answer, not per request.

Characters, not markets

Robust international support

A VIN is 17 characters wherever the vehicle was built, so extraction does not depend on the market. Decoding it afterwards does.

Customer stories

Companies love building with CarsXE

How teams build on CarsXE vehicle data.

“

CarsXE offers MotorTango's customers accurate and reliable vehicle data across many makes and models. Their VIN decoder and specs API are second to none.

Andy Liakos

CTO, MotorTango

“

The API is super easy to work with...it's a damn good API. And trust me, I deal with a lot of third parties and you're the creme de la creme. It's great. Documentation is sound, the result sets are sound. I have nothing to say but, man, it's too easy to work with.

Alan Deitch

Senior Engineering Leader, SpotHero

“

When a customer calls a dealership, we know they have an open recall instantly. That changes what the call is worth. Our agents book the recall alongside whatever the customer originally called about, and the dealer captures service work that used to walk out the door unnoticed. CarsXE's recall data is what makes that possible on every call, at every rooftop we run.

Samee Khan

Founder & CEO, PAM

“

With their VIN decoder and images, we can accurately document what's needed to perform repairs and ADAS calibrations, and we do it much more efficiently — saving dozens of minutes on every single repair. The result is consistent, reliable service at scale for all of our customers, backed by data we trust. CarsXE has become a critical part of our workflow and our ability to grow.

Philip Brossy

Founder & CEO, ADAS Safe

Same key, more endpoints

Keep building from the vehicle record

Vehicle Specifications

Full factory specs from a 17-character VIN

International VIN Decoder

Decode VINs from 50+ countries

Plate Image Recognition

Read a plate from a camera image

Vehicle Recalls

Open safety recalls by VIN

Lien & Theft Check

Open liens and theft records by VIN

FAQ

Frequently asked questions

Can't find the answer you're looking for? Reach out to our customer support team.

The best 17-character VIN read from the image, a confidence score between 0 and 1, a bounding box around the detected text, and ranked alternate candidates with their own scores.

Get started

Ready for the test drive? Start building today.

Test drive CarsXE's platform. Sign up with your work email.

Get startedLearn more
VIN OCR API