Vehicle Safety Report by VIN: Recalls, Complaints and NCAP Crash Ratings

One VIN in, one row out: what the US government knows about that vehicle's safety record. The VIN is decoded with NHTSA's official vPIC database, and the resulting make, model and year are checked against NHTSA's recall campaigns, its owner-complaint database, and its NCAP 5-star crash-test programme. No API key, no account, no proxy.

Who it's for

• Used-car buyers and marketplaces who need to know, before money changes hands, whether a car has an unrepaired recall on it, how often its model year has been complained about, and how it scored in a crash test.
• Dealers and wholesalers running a VIN list through a check before they list or bid on it.
• Fleet and insurance teams sweeping a whole fleet's VINs for a new do-not-drive recall.
• Anyone already using a VIN decoder who found that a decode alone answers the easy half of the question. The specs are on the window sticker; the recall history is not.

What you get

One flat row per VIN, never a nested blob you have to unpack:

What	Columns
The vehicle	vin, make, model, modelYear, manufacturer, vehicleType, bodyClass
Recalls	recallCount, recalls (every campaign in full), recallStatus
Owner complaints	complaintCount (NHTSA's own total), sampledComplaints, complaintStatus
Crash-test ratings	ncapOverallRating, ncapFrontCrashRating, ncapSideCrashRating, ncapRolloverRating, ncapVehicleId, ncapVehicleDescription, ncapModelMatched, ncapVariantsFound, safetyRatingStatus
Honest gaps	investigationsNote, sourcesChecked, sourcesFound

recalls is the full list, not a sample: each entry carries the NHTSA campaign number, the affected component, NHTSA's own summary, consequence and remedy text, the date the report was received, and the two do-not-drive flags (parkItAway, parkOutside).

sampledComplaints is a sample, and says so. A single common model year can carry thousands of complaints (a 2003 Honda Accord has 2,013 of them), so the row returns the most recently filed ones, newest first, capped at the number you choose. complaintCount is always NHTSA's own total for the vehicle, never the size of that sample.

What this is not

This is not a Carfax-style vehicle history report. It carries no odometer readings, no accident records, no title status, no branded-title or salvage check, no ownership or registration history, and no service records. Those come from state DMV records and commercial data suppliers, and this actor has no access to any of them. If that is what you need, buy a report from a provider licensed to sell one.

What this actor gives you is the other thing: official US federal safety data, free at source, for the vehicle as a model — recalls, owner complaints and crash-test ratings, keyed by the make, model and year your VIN decodes to.

Investigations are deliberately not included. NHTSA's Office of Defects Investigation publishes its open investigations only as a bulk flat file, not as a live route anything can query per vehicle. Rather than ship an always-empty investigations: [] that would read as "this vehicle has none", every row carries a short investigationsNote saying plainly that they are not covered here.

The four NHTSA sources

The VIN decode comes first and is not optional: NHTSA's recall, complaint and crash-test routes are all keyed by make, model and model year, never by VIN. The decode is what produces them.

1. vPIC VIN decode — resolves the VIN to a make, model, year, manufacturer, vehicle type and body style. If a VIN does not decode cleanly, it is a genuine miss: the other three sources have nothing to be looked up with, so they are never called and nothing is charged.
2. Recall campaigns — every NHTSA recall covering that make, model and year.
3. Owner complaints — the total NHTSA holds, plus your sample of the most recent ones.
4. NCAP crash-test ratings — the 5-star overall, frontal, side and rollover ratings.

Each of the last three is independent. One of them failing or returning nothing never fails the row and never affects the other two, and each one's own *Status column says exactly what happened.

A real subtlety in the crash-test data

NHTSA's crash-test catalogue does not use the same model names as its VIN decoder. A 2013 Ford F-150 decodes as model F-150, but NHTSA's crash-test programme lists that year's trucks as F-150 REGULAR CAB, F-150 SUPER CREW and F-150 SUPERCAB. A lookup that insisted on an exact name would report "no crash-test data" for a vehicle NHTSA plainly tested. So this actor tries the exact name first, and only if that comes back empty does it consult the catalogue for that year and make and match on the body-style-qualified names. The entry it actually used is reported as ncapModelMatched, so a star rating never arrives without its provenance.

NHTSA also often holds several separately tested variants for one make, model and year: 2-door versus 4-door, side-airbag versus not, rear-wheel drive versus all-wheel drive. A 2003 Honda Accord has four. This actor reports the first, and never averages them — averaging a 4-star coupe with a 5-star sedan would invent a rating NHTSA never gave to any car. ncapVariantsFound tells you how many exist and ncapVehicleDescription tells you which one the stars on your row belong to, so you can look the others up when the difference matters.

What a run costs, and how to control it

Billing is per source that answered, never one flat fee per VIN:

• A VIN that does not decode costs nothing at all. Nothing else was even looked up.
• A VIN that decodes charges the decode, plus one event for each of the other three sources that answered.
• Zero is an answer and it is charged. A vehicle with no recalls on record, no complaints on record, or no crash test on record has been checked, and the clean result is the thing you were asking for. Only a source that genuinely failed to respond, or one you switched off, is free.

The three Include checkboxes are the real cost control. Unticking "Include owner complaints" means the complaint route is never called and never billed for any VIN in the run. Hiding a column under "Which columns do you want?" is presentation only and does not change what a run costs; the input says so rather than implying a saving that is not there.

How to use Vehicle Safety Report by VIN

1. In the Apify Console. Open the actor page and click Start — the vins field is already pre-filled with a working example. Results land in the run's dataset as soon as each item is found.
2. Via the API. Call it directly with a POST request — no Console needed once you have an API token:

   curl "https://api.apify.com/v2/acts/accountable_eel~vehicle-safety-report-lookup/run-sync-get-dataset-items?token=<YOUR_TOKEN>" \
     -X POST \
     -H "Content-Type: application/json" \
     -d '{"vins":["1HGCM82633A004352","1FTFW1ET9DFC10312"]}'
   

3. On a schedule. Save this actor as an Apify Task with the input you want, then add a Schedule (hourly, daily, weekly) so it runs on its own — no server of your own required.

Input

{
  "vins": [
    "1HGCM82633A004352",
    "1FTFW1ET9DFC10312"
  ]
}

One 17-character VIN per line. Each VIN is decoded with NHTSA's official vPIC database, then its make, model and year are checked against NHTSA's recall, owner-complaint and NCAP crash-test records, and the whole safety record comes back as ONE row. Accepted formats: 1HGCM82633A004352, 1FTFW1ET9DFC10312.

Output

One row per VIN, for example:

query	found	status	vin	make	model	modelYear	manufacturer	vehicleType	bodyClass	recallCount	recalls	recallStatus	complaintCount	sampledComplaints	complaintStatus	ncapOverallRating	ncapFrontCrashRating	ncapSideCrashRating	ncapRolloverRating	ncapVehicleId	ncapVehicleDescription	ncapModelMatched	ncapVariantsFound	safetyRatingStatus	sourcesChecked	sourcesFound	scrapedAt
1HGCM82633A004352	True	OK	1HGCM82633A004352	HONDA	Accord	2003	AMERICAN HONDA MOTOR CO., INC.	PASSENGER CAR	Coupe	24	[{"campaignNumber": "19V182000", "component": "AIR BAGS:FRONTAL:DRIVER SIDE:INFLATOR MODULE", "summary": "Honda (American Honda Motor Co.) is recalling specific 2003 Acura 3.2CL, 2013-2016 ILX, 201...	OK	2013	[{"odiNumber": 11746949, "components": "AIR BAGS", "crash": true, "fire": false, "summary": "The vehicle was rear-ended on I-5 in Seattle Washington. The impact drove the subject car into the car ...	OK	Not Rated	Not Rated	Not Rated	Not Rated	4739	2003 Honda Accord 2-DR. w/SAB	Accord	4	OK	4	4	2026-09-30T23:50:58.457Z
1FTFW1ET9DFC10312	True	OK	1FTFW1ET9DFC10312	FORD	F-150	2013	FORD MOTOR COMPANY	TRUCK	Pickup	3	[{"campaignNumber": "19V075000", "component": "POWER TRAIN:AUTOMATIC TRANSMISSION:CONTROL MODULE (TCM/PCM/TECM)", "summary": "Ford Motor Company (Ford) is recalling certain 2011-2013 F-150 vehicles...	OK			REQUEST_FAILED	4	4	5	3	7340	2013 Ford F-150 Regular Cab PU/RC 4x4	F-150 REGULAR CAB	2	OK	4	3	2026-09-30T23:50:58.507Z

A VIN that does not decode comes back as a row with "found": false, a message carrying NHTSA's own reason (most often that the check digit in the 9th position does not compute), and no charge.

Example runs

Three inputs that need no credentials of any kind and return rows as they stand. Paste one into the Input tab and press Start.

1. The full safety record for two vehicles

{
  "vins": ["1HGCM82633A004352", "1FTFW1ET9DFC10312"]
}

2. Recalls only, for a fleet sweep

{
  "vins": ["1HGCM82633A004352", "1FTFW1ET9DFC10312"],
  "includeComplaints": false,
  "includeSafetyRatings": false
}

3. Deep complaint history for one vehicle

{
  "vins": ["1HGCM82633A004352"],
  "includeRecalls": false,
  "includeSafetyRatings": false,
  "maxComplaints": 100
}

Pricing

Pay-per-event, one event per source that answered. A VIN that does not decode charges nothing at all, and a source that fails to respond is free. Live per-event prices are shown on the actor page's pricing panel, which is always the authority.

Concretely: a VIN that decodes and gets answers from all three other sources is billed four events. A VIN you ran with complaints switched off is billed three. A VIN that does not decode is billed nothing.

Use it from Clay, n8n, Make, or an AI agent

This actor runs synchronously over plain HTTP — call it directly from a script, a workflow tool, or an AI agent, no Apify Console needed once you have an API token.

curl "https://api.apify.com/v2/acts/accountable_eel~vehicle-safety-report-lookup/run-sync-get-dataset-items?token=<YOUR_TOKEN>" \
  -X POST \
  -H "Content-Type: application/json" \
  -d '{"vins":["1HGCM82633A004352","1FTFW1ET9DFC10312"]}'

n8n. Add an HTTP Request node: Method POST, URL https://api.apify.com/v2/acts/accountable_eel~vehicle-safety-report-lookup/run-sync-get-dataset-items?token=<YOUR_TOKEN>, Body Content Type JSON, JSON Body {"vins":["1HGCM82633A004352","1FTFW1ET9DFC10312"]} (swap in an expression from an earlier node for a real value).

Clay. Add an "HTTP API" column: Method POST, URL https://api.apify.com/v2/acts/accountable_eel~vehicle-safety-report-lookup/run-sync-get-dataset-items?token=<YOUR_TOKEN>, Body {"vins":["{{VIN}}"]}, mapping the row's VIN into the vins array.

MCP. In Claude, Cursor, or any MCP client with the Apify MCP server, ask for "vin safety report: recalls complaints ncap rating" — the agent will find and run this actor.

vs. the alternatives

	This actor	A plain VIN decoder	A paid vehicle history report
Decoded make, model, year, body style	Yes	Yes	Yes
Recall campaigns with remedy text	Yes	No	Yes
Owner complaint history	Yes	No	Rarely
NCAP crash-test star ratings	Yes	No	Sometimes
Accident, odometer and title history	No	No	Yes
Cost per VIN	Cents	Cents	Dollars
API key needed	No	Usually no	Yes, and a licence

The honest summary: if you want to know whether a specific car was crashed and what its odometer said at each owner change, this is not that product. If you want to know whether the model is subject to an unrepaired recall, what owners keep reporting about it, and how it scored in a crash test, this returns all three in one row for a fraction of a cent.

Limits, stated plainly

• Recalls, complaints and ratings are per make/model/year, not per individual car. NHTSA's own APIs work that way. A recall campaign covering your model year covers your VIN only if your VIN falls in the affected range, and this data cannot tell you whether your car's recall was already repaired. For that, use NHTSA's own VIN-level recall page or a dealer.
• US vehicles only. NHTSA is a US federal agency, and its records cover vehicles sold in the US market. A VIN for a car never sold in the US may decode and still have nothing behind it.
• A VIN with a bad check digit is rejected. vPIC returns a make and model for many invalid VINs while flagging the check digit as wrong; those are treated as misses, not as vehicles, and are never charged. Most invented or mistyped VINs fail this way.
• Crash-test ratings are reported for the first tested variant, not averaged. See the subtlety above.
• "Not Rated" is NHTSA's wording, not a missing value. It means NHTSA tested the vehicle in some categories but did not issue a rating in that one, and it is passed through verbatim.
• Investigations are not included, for the reason given above.

Data & privacy

Everything this actor returns is public US federal government data, published by the National Highway Traffic Safety Administration and free at source. It reads nothing that is not already public, holds no personal data about vehicle owners, and does not look up registration, insurance or ownership records, which are not available to it.

One thing worth knowing: NHTSA's owner-complaint records sometimes include a partial VIN and a free-text narrative written by the complainant, which may mention a location or a dealer by name. Those fields come through as NHTSA published them. If you are redistributing rows, treat the complaint summaries as third-party text rather than as your own.

FAQ

Will this tell me if a recall on my car was already fixed? No. NHTSA's per-VIN recall-completion lookup is a separate, VIN-level service. This actor works at the make/model/year level, which tells you which campaigns could apply.

Why does a car with 2,013 complaints only show 20 of them? Because a dump of 2,013 complaint narratives in a spreadsheet cell is not a usable answer. complaintCount always reports the real total; raise "Complaints to sample per VIN" up to 100 if you need a deeper sample.

Why is ncapOverallRating "Not Rated" on a car that clearly has star ratings? NHTSA issues an overall rating only when it has run the full test battery. Older vehicles often have frontal and side ratings but no overall score. The individual ratings on the same row are the real answer for those.

Why does the model on my crash-test row not match the model in the decode? Because NHTSA's crash-test catalogue uses body-style-qualified names. ncapModelMatched shows the catalogue entry used; see the subtlety section above.

Do I need a proxy? No. All four routes are free, unauthenticated US government APIs that answer a plain request.

What happens to a VIN that isn't 17 characters? It is passed to NHTSA as given. vPIC attempts a partial decode and reports what it can; if the result is not a clean decode, the row is a free miss carrying NHTSA's own explanation.

Related actors

• NHTSA VIN Decoder — the decode on its own, when the specs are all you need.
• AutoScout24 Listing Lookup and Otomoto Listing Lookup — the listing side of a used-car check, for the European marketplaces.