AI Agents & MCP

Vehicle data for AI agents & MCP

Agents that answer vehicle questions need structured tools, not scraping. Wiring each REST endpoint by hand is brittle; a hosted MCP server gives agents native, discoverable tools with auth built in.

Start free See the API

Free plan includes 100 API calls/mo — no card required.

api.vehdb.com
# AI Agents & MCP — one call, real answers
GET /v1/cars?q=2024+RAV4+Hybrid
{ "data": [ {
  "uuid": "550e8400-e29b-…",
  "model": "RAV4", "mpg_combined": 39 } ] }
# Same UUID, the data you need
GET /v1/cars/{uuid}/recalls → same call your MCP tools make
GET MCP: recalls_for_vehicle(uuid) → native agent tool

Datasets that solve it

The exact data this segment leans on — all on the same UUID, all queryable by API or MCP.

Hosted MCP server: search_vehicles, get_vehicle, recalls_for_vehicle, tire_sizes_for_vehicle, my_lists

Full REST API with machine-readable OpenAPI 3.1 spec

Cars, motorcycles, boats & tire fitment on one schema

Same Sanctum token for REST and MCP; OAuth2 dynamic client registration

Base URL: https://api.vehdb.com/v1

What teams build

Real workflows ai agents & mcp ship on top of VehDB.

  • ChatGPT/Claude agents answering spec, recall and fitment questions
  • Agentic quoting & underwriting workflows
  • LLM apps grounded on cleaned EPA/NHTSA data
  • Zero-glue integrations via llms.txt & OpenAPI discovery

Why vehdb

The same advantages, whichever segment you sit in.

Enriched, not just listed

Specs carry EPA fuel economy on cars and live NHTSA recalls plus NCAP ratings — the next question is already answered inline.

Multi-asset on one UUID

Cars, motorcycles, boats and tire fitment share a model. The record your analyst sees is the record your backend queries.

MCP server built in

Expose VehDB as native tools to Claude, Cursor or any agent — no glue code, no scraping.

Transparent pricing

Quota-based plans, no per-seat surprises. The free tier is real: 100 calls/mo, no card.

Government-sourced provenance

Enrichment comes from U.S. federal public-domain data (EPA, NHTSA) — clean licensing, transparent origins.

Built for production

Stable UUIDs, JSON responses with pagination, and saved searches that notify you when new records match.

See the full provenance on the data sources page.

Other solutions

Same dataset, a different problem to solve.

Insurance & Underwriting

Rating and claims teams need consistent make/model/year specs, body class, fuel type and o...

Fleet & Telematics

Fleet platforms need to normalise mixed car, light-truck and powersport assets to one sche...

Dealers & Marketplaces

Listing and inventory tools need to auto-fill complete, consistent specs from a make/model...

Fintech & Lending

Loan origination and servicing need authoritative specs and recall/safety context keyed to...

Marketing & Audiences

Campaign and content teams need clean make/model/year taxonomies and spec attributes to se...

Parts & Aftermarket

Parts catalogs and e-commerce need accurate fitment — which tire sizes and configurations...

Compliance & Research

Compliance, safety and research teams need authoritative, government-sourced specs, recall...

Marine Insurance

Marine underwriters rate hulls on make, model, year, type and engine — yet there is no cle...

Boat Dealers & Marketplaces

Marine listing platforms inherit seller-typed make/model strings that break search and fil...

Marine Parts & Accessories

Marine parts catalogs need "does it fit this hull" answers keyed to make/model/year — the...

Tire Retail & E-commerce

Tire shops live and die by fitment accuracy: wrong-size returns are expensive. Maintaining...

Browse them all on the solutions overview.

AI-NATIVE · MCP SERVER

Your AI agent's vehicle-data tools —not glue code.

VehDB ships a native Model Context Protocol server. Claude, Cursor or any MCP client queries cars, recalls, fuel economy and tire fitment as first-class tools — no wrappers, no scraping, no glue code to maintain.

  • Native MCP tools — search_vehicles, tire_sizes_for_vehicle, recalls, my_lists — callable by Claude, Cursor or any agent.
  • OAuth2 for AI providers plus simple Bearer tokens. Same auth, same data as the REST API and workspace.
  • No scraping, no function-calling boilerplate, no schema drift — the agent gets typed, documented tools.
Explore the MCP server Connect your agent
claude_desktop_config.json 
{ "mcpServers": {
    "vehdb": { "url": "https://mcp.vehdb.com/mcp" }
  } }
# the agent calls a typed tool, not your scraper
tool search_vehicles(make="Toyota", year=2026)
tool tire_sizes_for_vehicle(uuid) → "235/65R17" …

Solve ai agents & mcp on VehDB.

Agents that answer vehicle questions need structured tools, not scraping. Wiring each REST endpoint by hand is brittle;... Start free, then scale by usage — not by seat.

Create free account Talk to sales

Operated by Savas OÜ (Tallinn, Estonia — EU) · GDPR-compliant · Billing by Stripe · Terms · Privacy