The idea

Vinted has no public API. This package bridges that gap — it lets AI assistants talk directly to Vinted via the Model Context Protocol.

Connect it to Claude, Cursor, or any MCP-compatible assistant and just ask:

"Find me a North Face jacket under €60 in Germany, good condition or better"

"Compare prices for Air Jordan 1s across France, Italy and the UK"

"What is seller #123456 currently selling? Anything under €20?"

The AI figures out which filters to use, calls Vinted, and gives you a real answer — no searching, no filtering, no tabs.

Also ships as a CLI tool and TypeScript library for direct use.

What is this?

An MCP server, CLI tool, and TypeScript library for the Vinted secondhand marketplace. No official API — it bootstraps a session cookie from the public catalog page and calls the private JSON API the Vinted web app uses internally.

• 🤖 MCP server — plug into Claude, Cursor, or any AI assistant with MCP support
• 🖥️ CLI tool — pipe results, watch for new listings, compare prices from your terminal
• 📦 TypeScript library — import opSearch, opCompare, etc. directly in your code

Install

npm install -g @googlarz/vinted-client

Or run without installing:

npx @googlarz/vinted-client search "levis 501"

CLI Quick Start

# Search (JSON by default)
vinted search "levi's 501" --country fr

# Pretty table
vinted search "levi's 501" --country de --output table

# Filter by price, brand, condition
vinted search "adidas samba" \
  --price-min 20 --price-max 80 \
  --brand adidas \
  --condition new_with_tags,very_good \
  --output table

# Watch for new listings every 30s
vinted search "air jordan 1" --watch 30

# Walk all pages and collect up to 500 results
vinted search "vintage denim" --all --max-items 500

# Get a specific item (ID or URL)
vinted item 1234567
vinted item https://www.vinted.fr/items/1234567

# Seller profile + active listings
vinted seller 987654
vinted seller-items 987654 --output table

# Cross-country price comparison (6 countries by default)
vinted compare "north face jacket" --output table

# Browse category tree
vinted categories --query shoes --output table

# Look up brand IDs
vinted brands "stone island"

# What's trending right now
vinted trending --country fr --output table

Commands

Global flags

Search flags

Supported Countries

fr de uk it es nl pl pt be at lt cz sk hu ro hr fi dk se

MCP Server

Drop Vinted into any MCP-compatible AI assistant (Claude, Cursor, etc.).

Setup — Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "vinted": {
      "command": "npx",
      "args": ["-y", "@googlarz/vinted-client/mcp"]
    }
  }
}

Setup — Claude Code

claude mcp add vinted -- npx -y @googlarz/vinted-client/mcp

MCP Tools

Example prompts once connected:

"Find me Nike Air Max 95s under €70 in Germany, size 43, very good condition"

"Compare prices for a North Face puffer jacket across France, Germany and Italy"

"Watch seller #987654 and tell me when they list something under €30"

Library Usage

import { VintedClient, opSearch, opCompare, opSearchAll } from '@googlarz/vinted-client';

const client = new VintedClient();

// Basic search
const results = await opSearch(client, {
  query: 'levi\'s 501',
  country: 'de',
  priceMax: 50,
  condition: ['very_good', 'good'],
  sortBy: 'price_low_to_high',
});

console.log(results.items);

// Collect all pages concurrently (3-page prefetch window)
const all = await opSearchAll(client, {
  query: 'vintage band tee',
  country: 'uk',
  maxItems: 300,
});

// Multi-country price comparison
const report = await opCompare(client, {
  query: 'air jordan 1 retro',
  countries: ['fr', 'de', 'uk', 'it'],
});

Client options

const client = new VintedClient({
  proxyUrl: 'http://proxy:8080',   // or VINTED_PROXY_URL env var
  cacheTtlMs: 60_000,              // response cache TTL (0 = disable)
  rateLimitPerSec: 3,              // requests/sec per country
  rateLimitBurst: 6,               // burst capacity
  timeoutMs: 20_000,               // per-request timeout
});

How it works

Vinted has no public API. This library:

1. Bootstraps a session by hitting vinted.{cc}/catalog and capturing the auth cookies the Vinted frontend sets.
2. Calls the private JSON API (/api/v2/...) with those cookies, mimicking browser request headers.
3. Re-bootstraps automatically on 401 — tokens expire, the library recovers silently.
4. Rate-limits per country with a token bucket (configurable burst + refill) to avoid 429s.
5. Caches responses with LRU+TTL — 60s for search results, 1h for static data like categories.
6. Falls back to HTML scraping for item pages blocked by DataDome (JSON-LD + regex extraction).
7. Prefetches 3 pages concurrently in opSearchAll to maximise throughput within the rate-limit budget.

Proxy support

If Vinted blocks your IP (common on cloud VMs and CI), set a proxy:

VINTED_PROXY_URL=http://user:pass@proxy:8080 vinted search "nike"
# or
vinted search "nike" --proxy http://user:pass@proxy:8080

Standard HTTPS_PROXY / HTTP_PROXY env vars are also respected.

Environment variables

Requirements

• Node.js ≥ 18
• Optional: playwright + puppeteer-extra-plugin-stealth for --browser / VINTED_BROWSER=1 mode

License

MIT © googlarz