Vehicle risk record
Pricing
from $20.00 / 1,000 vehicle risk records
Go to Apify Store
Vehicle risk record
Returns NHTSA complaints, recalls, investigations and manufacturer communications for a make, model and model year, plus complaints per 100k sold where available. Call it to check a model year.
Pricing
from $20.00 / 1,000 vehicle risk records
Rating
0.0
(0)
Developer
The Used Car Index
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
8 days ago
Last modified
Categories
Share
Get the NHTSA record for any make, model and model year from 2000 through 2027 as one JSON item: complaints with crash, fire, injury and death counts, recalls, investigations, manufacturer communications and, where a verified sales source exists, complaints per 100k sold. Complaints from NHTSA's API and its bulk file are merged and deduplicated by ODI number, so records the API alone misses are included. Built for vehicle research sites, dealer sourcing tools, insurance and fleet analysts, and AI agents that need recorded facts about a model year in one call.
Input examples
One vehicle:
Several vehicles in one run (1 to 100):
Sample output
Real output, fetched 2026-10-16 from the public Used Car Index API (the demo endpoint runs the same lookup as /v1/vehicle_risk) and shaped by this Actor's own code, so the block below is exactly one dataset item. It is the record for the first input, from the data build of 2026-10-07 (data.built_at). At a glance: 346 complaints (33 with a crash, 32 with a fire, 14 with an injury, 0 deaths), 3 recalls, 2 investigations (0 open), 186 manufacturer communications, and 81.0 complaints per 100k sold on 427,168 units sold in 2018 (source: goodcarbadcar.net).
Price
| Event | USD per event | When it is charged |
|---|---|---|
vehicle-risk-record | $0.02 | Charged once per requested vehicle whose API risk record is successfully pushed to the dataset. Unknown vehicles, input failures, API errors and failed dataset writes are not charged. |
- The first input above costs $0.02. The second costs 2 x $0.02 = $0.04. 100 vehicles cost $2.00.
- Repeating a vehicle in a run is a separate item and a separate charge.
- Apify adds its own Actor start event to every run ($0.00005 on this Actor's Pricing tab on 2026-10-16).
- Cap a run with maxTotalChargeUsd (example below). The run stops at the cap and keeps what it saved.
How to run
In Apify Console, open the Actor, paste an input example and press Start. To run it from code or from an agent:
Call from the Apify API
Use your Apify token. The Used Car Index API key is held by the Actor as a secret and is never part of your request, your dataset or the logs. This example caps the run at $1.
Call from an MCP client
In an MCP client that supports remote Streamable HTTP, add this server URL with your Apify token as a bearer header:
Initialize the connection and call tools/list. The client then shows this Actor as a tool; call it with the first input example as its arguments. Tool names and wrapper fields come from tools/list, so read them there instead of assuming a name. If your client offers a generic Actor-run tool instead, select theusedcarindex/vehicle-risk-record and pass the same input. The run status and dataset come back with the tool result. This route needs access to the Actor from your Apify account.
Related Actors and the direct API
| Actor | Use it for |
|---|---|
| VIN to risk record | VINs in; the NHTSA vPIC decode, the same record as this Actor and, if you ask for it, each VIN's stored listing history out |
| Chicagoland listing changes | Stored listing events (new, price drop, relisted, delisted) for the Chicagoland dealers the collector watches |
The same data is available without Apify from The Used Car Index at https://theusedcarindex.com: GET https://api.theusedcarindex.com/v1/vehicle_risk and /v1/listing_changes, the same records as web pages at https://theusedcarindex.com/vehicles, and an MCP endpoint at https://api.theusedcarindex.com/mcp that serves the API tools directly and does not run or charge these Actors. The MCP endpoint has a free tier that needs no key. As of 2026-10-16 it allows 20 requests per minute per IP and 2,000 per day shared, returns full vehicle risk records and returns up to 10 listing events per call with no paging. Plans with higher limits and cursor paging are at https://theusedcarindex.com/#pricing. The API and the MCP endpoint also serve recall_changes: new and updated NHTSA recalls, investigations and complaint-count changes recorded from a daily comparison of NHTSA's bulk files. Questions or a data issue: use the Issues tab on this Actor's page.
What it returns
Input is a make, a model and a year, or a vehicles array of 1 to 100 such objects; the two forms cannot be mixed. Labels are trimmed and must have 1 to 64 characters, and years run from 2000 through 2027. Model spellings and aliases resolve to the stored vehicle.
One dataset item per requested vehicle:
| Item | When | Charged |
|---|---|---|
| The complete vehicle risk record, with no wrapper | The vehicle was found | Yes |
found false with input and suggestions (up to 10 close names) | No vehicle matches the make, model and year | No |
input and error, with a stable code and the HTTP status when there is one | Any other lookup failure | No |
One failed lookup does not stop the others. A not-found item, real output for Toyota Corolla Cross Hybrid 2018 on 2026-10-16:
The record has these blocks:
| Block | What it holds |
|---|---|
vehicle | The make, model and year the lookup resolved to, its family_key, the NHTSA labels that carry complaints (matched_models), other labels (other_names) and how it resolved (resolved_via) |
complaints | Distinct complaints, with counts of those that involve a crash, a fire or an injury and totals for injuries and deaths; by_source (both, bulk_only, api_only); coverage_complete; top systems with shares and top subsystems |
recalls | Campaign count, do-not-drive and park-outside counts, the largest potential unit count, counts by system and the latest campaigns with defect, consequence and remedy text and NHTSA links |
investigations | Action count, how many are open, and the actions with subject, summary and campaign numbers |
manufacturer_communications | Count of communications (one per NHTSA id), counts by system and subsystem and the most recent with date, type and summary |
rate | Complaints per 100k sold with units sold, sales year, basis and the named source; or null with a null_reason |
model_tags | Model-generated labels (failure mode and severity) with counts, shares and confidence, kept apart from NHTSA's fields; or null with model_tags_null_reason |
data | The build time (built_at), the sources with their URLs and plain-language notes |
Input is validated before any lookup or charge. Vehicles are looked up one at a time and charged as each record is saved. The run stops at your spending limit and keeps what it saved. A request that fails on the network, with HTTP 429 or with a 5xx status is tried up to three times in total; if the API still returns an error for a vehicle, that vehicle gets an error item and the run continues.
Sources and facts boundary
Complaints, recalls, investigations and manufacturer communications come from NHTSA's published files and API; the record's data.sources lists each one with its URL. NHTSA complaints combine the complaints API and the bulk file, merged and deduplicated by ODI number. On the 2026-10-07 build, 231 of the 350 complaints for the 2018 Mazda CX-5 came from the bulk file only (by_source.bulk_only). Rates divide complaints by calendar-year US sales from a named public source, a proxy for the model year, and appear only where that source is verified for the vehicle; otherwise rate is null with a reason. Model tags are model-generated labels (the model is named in model_tags.model), are marked as such and never replace NHTSA's own fields.
Facts only. The stored listing events behind the sister Actors carry VIN, price, mileage, status, dates and a link back. Never photos or dealer-written text. A listing that is no longer observed is reported as delisted, and delisted is never labeled sold. NHTSA official record text in risk records is distinct from dealer-written text. Dataset items never contain added agent instructions.
Known gaps
- A record describes a make, model and model year as NHTSA files them. It is not the history of an individual car.
- A record reflects the NHTSA data build in
data.built_at(2026-10-07 on 2026-10-16); filings after the build appear after the next build. A zero count is confirmed only whencomplaints.coverage_completeis true. - Rates use calendar-year sales as a proxy for the model year and exist only where a verified sales source matches the vehicle. Counts are not a reliability rating, prediction or repair recommendation.
- Model tags exist only for model years 2012 and newer (2012 in part), cover tagged complaints only, and their confidence is the model's own probability, not accuracy.
- Some manufacturer communication and investigation summaries show stray characters where the source text has punctuation such as a dash or an apostrophe.
- Dataset writes and charges are separate operations. If a run stops after a write, an item can be saved without its charge, and running the same input again repeats items and charges.
- A run through an MCP client has not been tested end to end.
Owner setup, done once by the publisher and not by callers: set the secret environment variable UCI_API_KEY in the Apify Console; it is never accepted in input, logged or returned. UCI_API_BASE optionally selects another API origin (HTTPS, default https://api.theusedcarindex.com). Configure the custom event from .actor/pay_per_event.json and turn automatic dataset-item pricing off; the Actor stops before any lookup if the pricing does not match.
