Zero KM · MCP setup

Connect your AI agent

Give your agent tools to find buying opportunities and compare current new-car prices in Argentina.

API v1.0.0 No authentication Read-only

Connect your client

Add a remote MCP server in your client using the endpoint below. Choose Streamable HTTP and no authentication. Hosted clients need a public HTTPS address; localhost works for clients running on this machine.

MCP endpoint

https://zero-km.vercel.app/mcp

Codex setup

Terminal

codex mcp add zero-km --url https://zero-km.vercel.app/mcp
codex mcp list

OpenAI MCP setup

Available tools

Pricing tools return the same envelope as the HTTP API in structuredContent and a JSON text fallback. Pass numbers and booleans as JSON values; compare_vehicles.ids is an array. Errors set isError=true. Resources expose the developer guide and methodology as Markdown.

Tool	Use it for
`find_deals`	Find new-car buying opportunities in Argentina by budget, model and specs: below-usual prices, price drops in the last 1/3/7 days, or newly qualifying deals. Returns reasons, before/after prices where available, quote dates and links to explore on Zero KM.
`get_model_prices`	Get current prices for an exact make and model in Argentina: starting price, cheapest version and quote, price range, opportunity count and versions ordered by price. Follow the model link for charts and detailed exploration.
`search_vehicles`	Find the lowest current new-car prices in Argentina by budget, make, model and sourced specifications. Defaults to cheapest first. For recent price drops and buying opportunities use find_deals; get_vehicle provides all current quotes.
`get_vehicle`	Inspect a version’s current dealer cash quotes in original ARS/USD, fees, stock at observation, quote-check dates and official specification sources. Includes compact opportunity evidence and a Zero KM link for the full chart and saved shortlist.
`compare_vehicles`	Compare two to four versions by current prices, buying signals and sourced specifications. Returns a shareable Zero KM comparison page. Missing specifications remain unknown.
`get_catalogue`	Discover supported makes, models, filter values, coverage counts and snapshot dates before narrowing a price or opportunity search.
`search` / `fetch`	Retrieval-compatible discovery by make, model or version words, and a sourced vehicle record by ID.

Tool parameters and response schemas

Try these questions

“Which automatic SUVs under US$30,000 dropped in price in Argentina in the last seven days? Show the before/after prices and links.”
“What is the lowest current price for a Toyota Corolla, and which versions are below their usual price?”
“Compare these two versions by price and equipment, with links to inspect them on Zero KM.”

find_deals arguments

{
  "kind": "price_drops",
  "window": "7d",
  "type": "SUV",
  "transmission": "Automatic",
  "maxPriceUsd": 30000,
  "limit": 5
}

Make answers useful to buyers

Return the Zero KM url, original sourceUrl and individual quote observedAt. Link to the model or comparison page when that better matches the question.
Keep original currencies and quote conditions. Distinguish a recent price reduction from a usual-price discount; neither implies a manufacturer promotion.
Cite specification sources through specs.sourceOf and specs.sources. Missing equipment remains unknown.

Data rules

Check the connection

MCP Inspector should list 8 tools and two Markdown resources. The server uses stateless Streamable HTTP with current and 2025 client support. A plain browser GET to the endpoint can return 405; connect with an MCP client instead.

Terminal

npx @modelcontextprotocol/inspector