Changelog | VehDB

October 27, 2026

Added

New York DMV data on car pages: how many of the model year New York inspected in the last 12 months, the miles a year their owners drive (from odometer readings at inspection), the inspection fail rates against the model year's average, and the median curb weight. Boat pages show how many boats of the builder New York has registered by model year. GET /v1/cars/{uuid} returns them as new_york, GET /v1/boats/{uuid} as new_york_registrations, and get_vehicle carries both.
The VIN Decoder & Check shows the VIN's own New York inspection record: every inspection since August 2021 with its odometer reading and results, the miles a year between readings, and a note when a reading is lower than an earlier one. GET /v1/vin/{vin} and decode_vin return it as ny_inspections.
Washington plug-in prices: electric and plug-in hybrid car pages show what the model year sold for new and used in Washington by year of sale, how much of the new price a used one kept, and how many are registered there. GET /v1/cars/{uuid} returns it as washington_ev. The aggregates are free to download under the Open Database License at /data/washington-ev-prices.

October 20, 2026

Added

Boat history by HIN: GET /v1/hin/{hin}, the Boat HIN Lookup (/hin-lookup) and the decode_hin MCP tool (with include_history) list the recreational boating accidents reported under a hull from 1995 to 2023, from the U.S. Coast Guard accident reports released under the Freedom of Information Act (date, state, events, damage, complete loss, people hurt), and its record in the Coast Guard's monthly file of federally documented vessels, with the name, hailing-port and certificate changes seen month to month. Every response says what the sources cover: no record found is not a clean record.

October 17, 2026

Added

Service bulletins and warranty extensions: car and motorcycle pages list the NHTSA manufacturer communications for the model year (technical service bulletins, special coverage and warranty extensions, over-the-air updates; NHTSA's files go back to 1995). GET /v1/cars/{uuid}/tsbs, GET /v1/motorcycles/{uuid}/tsbs and the service_bulletins_for_vehicle MCP tool return them. A bulletin is not a recall.
U.S. Coast Guard data for boats: boat pages list the recall campaigns filed under the make's manufacturer identification codes (MIC, the first three characters of a HIN), this model year first and the campaigns naming the model flagged, together with the codes themselves. GET /v1/boats/{uuid}/recalls returns them, and recalls_for_vehicle now takes a boat uuid. GET /v1/hin/{hin} and the decode_hin MCP tool decode a hull identification number: builder, build month and year, model year, the builder's recalls and the catalogue makes under its code.
Tire recalls and the DOT check: production windows are read from the text of NHTSA tire recall notices. The free Tire DOT Code Lookup (/tire-dot-check), GET /v1/tire-sizes/dot/{tin} and the tire_dot_check MCP tool read a DOT code into plant, build week and age, and list the tire recalls covering that plant and week. Tire size pages list the recalls naming their OEM size, also at GET /v1/tire-sizes/{uuid}/recalls.
Energy prices from the U.S. Energy Information Administration: GET /v1/energy-prices?state= and the energy_prices MCP tool return the latest weekly retail gasoline (by grade) and diesel prices for a state, its PADD region or the U.S., and monthly electricity prices by state. Car pages show the fuel cost of the EPA rating at this week's pump price.
Motorcycle engine specs from EPA emissions certification (highway and off-road): displacement, cylinders, rated power in kW and hp at rpm, transmission and gears, on the records and pages of models a certified engine family matches. Motorcycle pages also show NHTSA owner complaints.
Four new MCP tools, 16 in all: service_bulletins_for_vehicle, decode_hin, tire_dot_check and energy_prices.
Gas vs Electric Cost Calculator (/fuel-cost-calculator): two vehicles' EPA consumption against every state's latest EIA gasoline, diesel and electricity price, ranked state by state. Car pages carry a running-cost card with the same prices.
Boat HIN Lookup (/hin-lookup): decode a boat's Hull Identification Number to its builder, build month and model year, with the Coast Guard recalls filed for the builder.

October 17, 2026

Improved

The VIN decoder is now the VIN Decoder & Check: after the decode it shows the model year's report, with recalls, the complaint ranking and its best and worst year, service bulletins and warranty extensions, fatal crashes, fuel cost at EIA prices and tire sizes.
The Tire Finder takes a VIN as well as make, model and year.
Recall watch covers boats: listed boats are checked against U.S. Coast Guard campaigns for their builder and model year on the dashboard, in GET /v1/lists/recalls and in the daily recall digest.
The model-year scorecard counts service bulletins and warranty extensions per year and flags years with special coverage.

October 3, 2026

Added

AI search: when a search finds nothing, VehDB suggests the closest models by meaning or spelling ("family minivan", "cambry" → Toyota Camry). In the API, list endpoints and /v1/resolve return them as suggestions; add ai=true for their records.

October 1, 2026

Changed

GET /v1/{type}/models?make= now answers on the Free plan with model names only (limited: true), instead of a 403, so a free key can drill down from makes to models to records. Year ranges and record counts per model stay on Pro and higher.

September 26, 2026

Added

One-click AI connectors: VehDB is listed in Claude's connector directory (claude.ai/directory/vehdb). Click Connect and approve with your VehDB login; no server URL to paste. Calls share your team's plan quota, as with every other client.

September 26, 2026

Added

Markets: the catalogue lists US, Canadian and Mexican models, now named on every record (markets: ["US", "Canada"]) and filterable with market=us, ca or mx on GET /v1/cars. Search in the dashboard and the search_vehicles MCP tool default to models sold in the US, with a switch to the other markets.

September 26, 2026

Improved

EPA fuel economy for models EPA spells differently: Tesla ("Model 3"), Lexus ("ES 350"), Mercedes-AMG and Maybach, Mazda MX-5 Miata and more. EVs carry MPGe and EPA range per version (Long Range, Performance, Plaid), and a version EPA did not rate gets no figure rather than a neighbour's.
NHTSA vPIC model ids now cover Tesla, the Mercedes-Benz classes and Mazda, so VIN decodes and model links reach those vehicles.
Leaner MCP answers: compact JSON, vehicle records without empty fields or internal ids, and recalls_for_vehicle listing the newest 10 campaigns (limit up to 50) with short summaries and a link to each campaign on nhtsa.gov. Recall answers are about 75% smaller and search results about half; recalls are metered by the campaigns listed.

September 26, 2026

Added

PDF spec sheets on paid plans: one click on a car page downloads its specifications, EPA fuel economy, NHTSA safety (rating, recall campaigns, open investigations) and OEM tire sizes as a one-page PDF.

September 26, 2026

Improved

Versions of a trim are told apart everywhere: search results show engine, drive, transmission, body and MPG, and mark the ones that differ between rows of the same trim; a car page names its version ("Camry 2025 XLE · AWD") and links the others.
Similar vehicles are rival models of the same body type, size class and powertrain from adjacent model years, each marked with what differs from the car you are looking at and its MPG difference.

September 25, 2026

Added

BMW series and Mercedes-Benz classes: model=3 Series, 3er or C-Class on the list endpoints, Discovery and the search_vehicles MCP tool returns the member models (330i, M340i; C300, C43 AMG), and free-text "bmw 3 series" or "mercedes c class" finds them. GET /v1/resolve and resolve_vehicle accept these names and list the members in a new catalogue_models field.
Crash records carry crash_match {status, models}: the FARS and CRSS names a vehicle was matched under, like nhtsa_match for recalls. Status "none" means no crash record carries the name, not zero crashes.

September 25, 2026

Improved

Recalls and complaints NHTSA files under a whole BMW series or Mercedes-Benz class ("3 SERIES", "C-CLASS") apply to every model in it.
Crash data (GET /v1/cars/{uuid}/crashes, the crash_severity report and car pages) reads the spellings FARS and CRSS use, such as Tesla "Model S" and Mercedes-Benz "C-Class".
Model-year scorecards and make pages count recalls, complaints and fatal crashes under every spelling of a model, count each recall campaign once, and show EPA production volume for more models.

September 24, 2026

Added

Name resolution: GET /v1/resolve?q=2019+chevy+silverado+1500 returns the catalogue spelling of a typed vehicle (make, model, catalogue_match) and, with a model year, the NHTSA spellings its recalls and complaints are filed under (nhtsa_match). Also the resolve_vehicle MCP tool. Counts one row.
Search reads the names people type: "tesla model 3", "vw golf" and "chevy silverado" find the catalogue's Tesla 3, Volkswagen and Chevrolet, and the make/model filters on the list endpoints accept make=Chevy, Tesla model=Model S or F150.

September 24, 2026

Changed

Recalls, owner complaints and investigations are matched under every spelling NHTSA files a vehicle by (Tesla "S" is "MODEL S", a 2012 Ram 1500 is Dodge "RAM 1500", F-250 Super Duty is "F-250 SD"). Recall and complaint responses (the safety block on records, /recalls, /complaints, /vin/{vin}/recalls, /lists/recalls and the MCP tools) add nhtsa_match {status, models}: status "none" means no NHTSA record carries the name, so a recall_count of 0 there means "not matched" rather than "no recalls".
Open investigations are listed once each, with the recall campaigns they link in campaigns.
Recalls by VIN: vPIC's family name finds NHTSA's numbered models (Silverado → SILVERADO 1500 and SILVERADO), narrowed to the VIN's weight class. Decoded VINs include series.
VIN decodes are metered by what they return: GET /v1/vin/{vin}, POST /v1/vin/batch and the decode_vin MCP tool count one row for the decode plus one per matched VehDB record (up to 10 cars and 10 tire fitments, or 10 motorcycles), instead of one row per VIN.
A per_page above your plan's page size (10 Free, 25 Pro/Developer, 100 Enterprise) is clamped instead of rejected with a 422, on the list endpoints and GET /v1/tire-sizes/vehicles. Every page carries an X-Per-Page-Limit header; a clamped page adds meta.per_page_requested, meta.per_page_limit and meta.per_page_notice. A list query that does not narrow the dataset gets a working example request back in its 422.
On the list endpoints q narrows a search on its own when it holds a word of at least 2 letters or digits (q=X5 works); next to make or model any q refines.
MCP tool calls that end in an error result (an unknown uuid, a missing argument, an NHTSA outage) meter no records and do not count against the monthly call quota, like failed REST calls.
The crash_severity report of the vehicle_insights MCP tool counts one record plus one per CRSS severity row, like GET /v1/cars/{uuid}/crashes.
GET /v1/insights/recalls/brands and the MCP recall_brands report now spell out what units means: the vehicles in every campaign involving the brand. A campaign filed for several brands (Chrysler, Dodge, Jeep and Ram share many) counts in full for each, so units are not additive across brands. A new totals block carries the year's distinct campaigns, brands, vehicles and do-not-drive campaigns, and a note explains the overlap. The per-brand values are unchanged.

September 23, 2026

Added

Recalls by VIN: GET /v1/vin/{vin}/recalls decodes the VIN and returns the NHTSA recall campaigns for its make, model and year, with manufacturer campaign numbers, build windows and engine hints. It is not a per-VIN repair check (match_level make_model_year, repair_status null, disclaimer on every response). The recalls_for_vehicle MCP tool takes a vin, and the recall checker has a VIN tab.

September 23, 2026

Changed

Motorcycle complaints: GET /v1/motorcycles/{uuid}/complaints returns NHTSA owner-complaint statistics in the same shape as the car endpoint (counts only, no narratives). The recalls_for_vehicle MCP tool now accepts motorcycle uuids.
New MCP tools matching the REST API: decode_vin (GET /v1/vin/{vin}), vehicles_for_tire_size (GET /v1/tire-sizes/vehicles) and api_usage (GET /v1/me). search_vehicles, tire_sizes_for_vehicle and my_lists take a page argument, return results in a stable order and treat % and _ literally.
VIN matches: make is matched exactly and model exactly or by prefix; motorcycle VINs return matches.motorcycles. Batch decodes run concurrently.
GET /v1/insights/recalls/brands adds units (vehicles covered, each campaign counted once) and ranks ties by units; it is now the same report as the most-recalled-brands page. /v1/insights/ev-share uses the same electrified definition as ev_only=1.
Every response carrying tire data (list and single fitments, geometry, rim-size trend, VIN matches, and the matching MCP tools) includes the fitment disclaimer.
GET /v1/lists is paginated (max 100 per page, meta block like /v1/lists/recalls). GET /v1/cars/{uuid}/tire-sizes returns at most 50 fitments, nearest model years first.
Row metering now also covers recalls, car tire sizes, tire geometry, single VIN decodes and lists; each served record counts against the monthly data limit.
API errors are JSON on every host whatever the Accept header: {"message": …}, plus errors on 422; unknown uuids and paths return 404 {"message": "Not found."}.

September 9, 2026

Added

Crash record: GET /v1/cars/{uuid}/crashes returns NHTSA FARS (fatal crashes, census, exact) and NHTSA CRSS (all severities, weighted sample, national estimates) side by side and never mixed; the scorecard and car pages show both, labelled by source.
EPA test data: GET /v1/cars/{uuid}/epa-tests lists the dynamometer configurations behind the label — rated horsepower, test weight, axle ratio, gears and raw city/highway results. Cars now carry smog_rating and smartway from the EPA Green Vehicle Guide, and the scorecard shows EPA CAFE production volume where published.
Tires: GET /v1/tire-sizes/plants/{code} decodes the DOT plant code on a sidewall; geometry responses and the tire calculator decode load index and speed symbol into kg/lb and km/h/mph. Tire surfaces carry a fitment disclaimer.
Model year scorecard: GET /v1/insights/scorecard?make&model lines up every dataset per model year — trims and EPA mpg, NHTSA complaints with a percentile against the same model-year cohort, crash and fire share, top component, recalls and do-not-drive advisories, open investigations, NCAP stars and FARS fatal-crash involvement — with best/worst year and flags. Also the model_year_scorecard MCP tool and the free /model-year-scorecard page.
Insights endpoints: /v1/insights/complaints/vehicles and /complaints/components, /v1/insights/fatal-crashes/vehicles and /body-classes (occupant deaths per involved vehicle), /v1/insights/engines (displacement, cylinders, turbo share by year), /v1/insights/fuel-economy/canada (NRCan), /v1/insights/recalls/age (recall timing by vehicle age) and /v1/insights/tires/rim-sizes — and the vehicle_insights MCP tool that wraps them all.
Insights pages: most complained-about vehicles per model year, and vehicles in fatal crashes per year with survivability by body class.
Owner complaints: GET /v1/cars/{uuid}/complaints returns NHTSA ODI complaint statistics for a make/model/year — totals with crash, fire, injury and death counts, the most reported components, complaints by year received, and neighbouring model years. Aggregated weekly from the full NHTSA complaint file (2.2M records); counts only, no narratives.
Car pages show an Owner complaints panel: complaints by model year, when owners complained, and the most reported components — the same data as the endpoint.
MCP tools complaints_for_vehicle (owner-complaint statistics) and recalls_for_vehicle (live recall campaigns, open investigations and NCAP rating, by uuid or make/model/year).
Shop parts on car pages: eBay Motors and Amazon searches for ten common part categories, pre-filled with the year, make, model and engine size.
New public-data mirrors behind the scenes: NHTSA FARS fatal-crash vehicles (2020–2024), Natural Resources Canada fuel consumption ratings (1995–2026) and NHTSA vPIC make, model and engine ids linked to the catalogue. Surfaces for these follow.

September 2, 2026

Added

Fleet recall monitor: a daily email digest when NHTSA publishes a new recall for any car or motorcycle saved on your lists, and GET /v1/lists/recalls to see recall exposure across every saved vehicle in one call.
POST /v1/vin/batch — decode up to 100 VINs per request, each with its own status and matching VehDB uuids.
Inline safety block on GET /v1/cars/{uuid} and /v1/motorcycles/{uuid}: recall_count, do_not_drive, park_outside, open_investigations and ncap_overall from the NHTSA mirror.
GET /v1/tire-sizes/vehicles?size=225/60R17 — reverse fitment: every vehicle that runs a size as OEM or alternate.
GET /v1/boats/{uuid}/similar — comparable vessels by category, length and horsepower; boat list filters now also accept beam, weight, lb_per_hp and length_beam_ratio ranges.
Insights endpoints: GET /v1/insights/recalls/brands, /v1/insights/fuel-economy and /v1/insights/ev-share — aggregate views for analysts and AI agents, no record sweeps needed.

July 16, 2026

Added

Boat spec filters on GET /v1/boats: hull_type, fuel_type, model_type, hp_min/hp_max and length_min/length_max — combine them with make, model, year or q to narrow marine searches.
Boat taxonomy endpoints: GET /v1/boats/makes (every manufacturer with vessel count) and GET /v1/boats/hull-types (every hull type with vessel count).
Car spec filters on GET /v1/cars: body, fuel_type, drive_type and ev_only=1 (BEV/PHEV/HEV) — same semantics as the dashboard filters.
Car taxonomy endpoints: GET /v1/cars/makes and GET /v1/cars/body-types, each with record counts.

July 15, 2026

Added

GET /v1/vin/{vin} — decode any 17-character VIN via NHTSA vPIC and get matching VehDB car and tire-fitment uuids in the same response, ready for follow-up spec, recall and geometry calls. Same decoder that powers the free /vin-decoder tool.

July 2026

Added

VehDB Insights: free data-journalism reports on the official NHTSA dataset, starting with the most recalled car brands ranked per year (2010 to today).
Interactive API reference at /docs/api: every endpoint, parameter and schema, with runnable example requests, generated live from the OpenAPI spec.
The full NHTSA recall dataset is now mirrored into VehDB and refreshed weekly; EPA fuel-economy enrichment refreshes monthly.

July 2026

Improved

Search understands multiple words: "toyota camry" now finds the Camry instead of nothing. Word-based, index-backed matching across Discovery, the dashboard typeahead and the API.

June 2026

Added

Vehicle Data by Region — North America is live today; Europe & UK and Australia & NZ are in early access as those datasets come online.
A growing developer reference: a vehicle-data glossary, provider comparisons, and a complete "vehicle data for business" guide.
An llms.txt index so AI agents and assistants can discover the API and MCP server directly.

June 2026

Improved

Faster, more reliable page loads — self-hosted fonts and icons, optimized and long-cached assets, and Core Web Vitals tuning across the marketing site.

June 12, 2026

Changed

API tokens are now team-scoped (Stripe-style account keys): tokens are created and managed per team, and quota/rate limits follow the team's plan.
GET /v1/me now returns the acting team's name alongside plan and quota.
GET /v1/lists returns the acting team's lists.

June 12, 2026

Added

GET /v1/cars/{uuid}/recalls — live NHTSA recall campaigns + NCAP crash-test ratings per car.
GET /v1/motorcycles/{uuid}/recalls — NHTSA recall campaigns for motorcycles.
GET /v1/tire-sizes/{uuid}/geometry — computed OEM size geometry and alternate-size diameter deltas (±3% fitment window).
EPA enrichment on car records: mpg_city, mpg_highway, mpg_combined, epa_range_miles, annual_fuel_cost, co2_gpm (null where EPA has no data).
Machine-readable OpenAPI 3.1 description at /openapi.json.

June 11, 2026

Added

MCP server at mcp.vehdb.com/mcp — query vehicle data natively from Claude, Cursor and other AI agents (path fallback /mcp/vehicles remains).
API vanity domain api.vehdb.com/v1 (path fallback /api/v1 remains).
MCP tools: search_vehicles, get_vehicle, my_lists, tire_sizes_for_vehicle.
OAuth2 (dynamic client registration) for AI providers, alongside Bearer tokens.
GET /v1/cars/{uuid}/tire-sizes — car → tire fitment cross-reference.
GET /v1/lists — the authenticated user's curated vehicle list.

March 2026

Changed

A redesigned VehDB — a cleaner, faster workspace and a refreshed brand across the whole platform.
Search and discovery reworked around a single, consistent results experience.

February 2026

Added

My Garage — organize saved vehicles into owned, wishlist and considering, with notes.
Tire Finder and Reverse Tire Search — find OEM and compatible sizes, or work backwards from a size to the vehicles that fit it.
Model Year Diff — compare how a model changed between two years.
EV Transition Advisor — surface electric and hybrid alternatives to a given vehicle.
VIN Decoder for quick make/model/year lookups.

December 2025

Added

Accounts and subscriptions — Free, Pro, Developer and Enterprise plans with secure hosted checkout.
Save vehicles to favorites with private notes.

July 2025

Improved

Better discovery — public catalog pages and structured data so search engines (and people) can find the right vehicle faster.

June 2025

Launched

VehDB is live — a searchable database of cars, motorcycles, power boats and tire sizes.
Advanced multi-attribute search and filtering, side-by-side comparison, and detailed specification pages for every record.