Agents & AI
Everything you need to use CarsXE vehicle data APIs from AI coding agents, LLMs, and automated pipelines — including browser-based onboarding, machine-readable documentation, and content negotiation.
CarsXE is built to be agent-friendly. Whether you are a coding agent onboarding a new project, an LLM reasoning about vehicle data, or an automation pipeline calling REST endpoints, this page covers the features we ship specifically for AI and agent use cases.
Browser-based agent onboarding
Getting a CarsXE API key from inside any AI agent — Claude Code, Claude Desktop, Cursor, Codex CLI, Gemini CLI, or any other coding assistant — takes just a few steps with no CLI install required. The agent directs you to a browser page, waits for you to authorize, then retrieves and stores your key automatically.
The complete machine-readable skill document for agents is available at
https://api.carsxe.com/agent-onboarding/SKILL.md. Agents that support skill discovery (such as Claude Code) will
pick this up automatically.
Step 1 — Give your agent the onboarding skill
Pass the skill document URL to your agent so it knows exactly how to authenticate:
"Use this skill to get a CarsXE API key: https://api.carsxe.com/agent-onboarding/SKILL.md"
The agent fetches the skill, generates secure session parameters, then presents you with a clickable link to open in your browser.
Step 2 — Authorize in the browser
Click the link the agent provides. You will be taken to the CarsXE authorization page where you can sign in or create an account and approve access. No typing required — just click Authorize.
Step 3 — Your agent retrieves the key
Once you click Authorize, the agent polls in the background and retrieves your API key. It stores it automatically (typically as CARSXE_API_KEY in your environment or project .env file).
Step 4 — Start making calls
Your agent can immediately start making vehicle data calls on your behalf:
"Decode VIN WBAFR7C57CC811956" "What is the market value of this vehicle with 50,000 miles?" "Check for any open recalls."
Machine-readable documentation (Accept: text/markdown)
CarsXE documentation pages support HTTP content negotiation. Send Accept: text/markdown on any documentation, support, or guide page to receive clean Markdown instead of HTML — fewer tokens, no layout markup, and higher signal for RAG pipelines and LLM context windows.
How to use it
Terminal
# Any docs page as Markdown
curl -H "Accept: text/markdown" https://api.carsxe.com/docs/quickstart
# Specific API reference page
curl -H "Accept: text/markdown" https://api.carsxe.com/docs/v1/specifications
# Direct Markdown endpoint — no content negotiation needed
curl https://api.carsxe.com/api/markdown/docs/v1/specificationsSupported content types
The site produces two content types:
| Type | Description |
|---|---|
text/html | Default browser response |
text/markdown | Clean Markdown — no layout, no navigation chrome |
If the Accept header explicitly excludes both (for example Accept: application/pdf), the server returns 406 Not Acceptable with a plain-text body listing the available types.
The Vary: Accept header
All responses include Vary: Accept so HTTP caches store separate entries for the HTML and Markdown representations of the same URL.
The Link: rel="alternate" header
HTML responses on documentation, support, and guide pages include a Link header advertising the direct Markdown URL. Agents that inspect response headers can discover the Markdown endpoint without making a second request.
Response Headers
Link: </api/markdown/docs/v1/specifications>; rel="alternate"; type="text/markdown"
Vary: AcceptDirect Markdown endpoint
The /api/markdown/[[...slug]] route serves text/markdown unconditionally — no Accept header required. Use it when you want to hard-code the Markdown URL in a retrieval pipeline or a prompt template.
Terminal
curl https://api.carsxe.com/api/markdown/docs/quickstart
curl https://api.carsxe.com/api/markdown/docs/authentication
curl https://api.carsxe.com/api/markdown/docs/v1/specifications
curl https://api.carsxe.com/api/markdown/docs/v2/market-value
curl https://api.carsxe.com/api/markdown/support/ARTICLE_SLUG
curl https://api.carsxe.com/api/markdown/guides/GUIDE_SLUGResponses include Cache-Control: s-maxage=300, stale-while-revalidate=86400 and Vary: Accept.
Site index for LLMs (llms.txt)
The site exposes a machine-readable index at /llms.txt. This document provides a plain-text overview of CarsXE for language models — product description, all API endpoints with pricing, subscription tiers, quick links, and recent blog posts. It is generated on each request with live pricing data so models always see current overage rates and tier names.
Terminal
curl https://api.carsxe.com/llms.txtMaking API calls from an agent
All endpoints accept key=YOUR_API_KEY as a query parameter — the same environment variable the onboarding flow stores. POST endpoints also send the body as JSON with Content-Type: application/json.
Vehicle Specifications (VIN Decode)
Decode a VIN and retrieve full vehicle specifications: make, model, year, engine, trim, equipment, and more. See full reference.
Request
curl -G https://api.carsxe.com/specs \
-d key=YOUR_API_KEY \
-d vin=WBAFR7C57CC811956Market Value
Estimate retail, trade-in, and auction values for a vehicle by VIN, with optional mileage, state, and condition adjustments. See full reference.
Request
curl -G https://api.carsxe.com/v2/marketvalue \
-d key=YOUR_API_KEY \
-d vin=WBAFR7C57CC811956 \
-d mileage=50000 \
-d state=CA \
-d condition=cleanVehicle History
Retrieve a comprehensive history report including title records, junk/salvage events, and insurance information for a VIN. See full reference.
Request
curl -G https://api.carsxe.com/history \
-d key=YOUR_API_KEY \
-d vin=WBAFR7C57CC811956Vehicle Recalls
Check for open safety recalls associated with a VIN. See full reference.
Request
curl -G https://api.carsxe.com/v1/recalls \
-d key=YOUR_API_KEY \
-d vin=WBAFR7C57CC811956Lien & Theft Check
Check whether a vehicle has active liens or has been reported stolen. See full reference.
Request
curl -G https://api.carsxe.com/v1/lien-theft \
-d key=YOUR_API_KEY \
-d vin=WBAFR7C57CC811956Plate Decoder
Look up a vehicle by license plate number. Supports 50+ countries. See full reference.
Request
curl -G https://api.carsxe.com/v2/platedecoder \
-d key=YOUR_API_KEY \
-d plate=7XER187 \
-d country=US \
-d state=CAInternational VIN Decoder
Decode a non-US VIN from European, Asian, and other markets. See full reference.
Request
curl -G https://api.carsxe.com/v1/international-vin-decoder \
-d key=YOUR_API_KEY \
-d vin=WF0MXXGBWM8R43240Vehicle Images
Retrieve vehicle photos by make, model, and year. Supports filtering by color, angle, photo type, and size. See full reference.
Request
curl -G https://api.carsxe.com/images \
-d key=YOUR_API_KEY \
-d make=BMW \
-d model=5-Series \
-d year=2012Year / Make / Model
Look up vehicle data when you don't have a VIN — search by year, make, and model instead. See full reference.
Request
curl -G https://api.carsxe.com/v1/ymm \
-d key=YOUR_API_KEY \
-d year=2012 \
-d make=BMW \
-d model=5-SeriesPlate Image Recognition
Extract and decode a license plate number from an image URL. See full reference.
Request
curl -X POST "https://api.carsxe.com/platerecognition?key=YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"image_url": "https://imagedelivery.net/moyiiSImjJPI_EZVxNMBBw/f49aed53-d736-4370-f3f4-97418841c800/public"}'VIN OCR
Extract a VIN from a photo of a VIN plate or dashboard sticker. See full reference.
Request
curl -X POST "https://api.carsxe.com/v1/vinocr?key=YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"image_url": "https://user-images.githubusercontent.com/5663423/30922082-64edb4fa-a3a8-11e7-873e-3fbcdce8ea3a.png"}'OBD Codes Decoder
Decode a diagnostic trouble code (DTC) from an OBD-II scanner — over 3,000 codes supported. See full reference.
Request
curl -G https://api.carsxe.com/obdcodesdecoder \
-d key=YOUR_API_KEY \
-d code=P0300MCP server for AI tools
For interactive AI editors and chat clients (Claude Desktop, Cursor, VS Code, Windsurf), the CarsXE MCP server is the recommended integration path — it exposes all endpoints as named tools and formats responses as Markdown automatically.
See the MCP Quickstart Guide for installation instructions for each editor.
What's next?
- Agent onboarding skill document — machine-readable PKCE auth flow
- llms.txt — full site index for LLMs
- MCP Quickstart — MCP server for Claude, Cursor, VS Code, Windsurf
- Claude Code Plugin — slash commands and auto-invoked skills
- CLI Quickstart — terminal tool for agent-driven shell workflows
- Authentication — API key management
- Quickstart — first API call walkthrough