Build with BasketBooster

Docs for your AI assistant (MCP)

Claude Code, Cursor, and other MCP-capable assistants can connect directly to our documentation server. A connected assistant can list, read, and search every integration guide, so a prompt like "add BasketBooster recommendations to my product page" gets answered from the current docs rather than the model’s memory.

claude mcp add --transport http basketbooster-docs https://api.basketbooster.eu/mcp

{
  "mcpServers": {
    "basketbooster-docs": { "url": "https://api.basketbooster.eu/mcp" }
  }
}

{
  "mcpServers": {
    "basketbooster-docs": {
      "type": "http",
      "url": "https://api.basketbooster.eu/mcp"
    }
  }
}

The server exposes three tools: list_docs, read_doc, and search_docs. The guides cover getting started, widget embedding, the full API reference, the WooCommerce plugin, keys and CORS, and plans and billing. You don’t need an account or an API key to connect.

Plain-text docs (llms.txt)

No MCP? Every guide is also published as raw markdown, indexed at https://api.basketbooster.eu/llms.txt. Each file is plain markdown, so you can paste a URL into ChatGPT, Claude, or your editor and work from it directly.

Interactive API reference (Swagger)

Explore and try the public API in the browser at https://api.basketbooster.eu/docs. It covers the complete integration surface:

The raw OpenAPI spec lives at https://api.basketbooster.eu/openapi.json. Feed it to a code generator or to your AI assistant.

Conversion tracking and attribution

The <product-recs> widgets report product views (detail-page-view), recommendation impressions (rec-impression), and recommendation clicks (rec-click) automatically. No site code is needed for those. They cannot see your cart button or your checkout, though, so add-to-cart and purchase events have to come from your storefront. The WooCommerce plugin sends both for you; on a custom storefront you wire them up once.

<script>
  ProductRecs.track('add-to-cart', {
    item_id: '123',
    user_id: 'CUSTOMER_ID',  // logged-in customer id, omit for guests
    value: 49.9,             // line value (unit price x quantity)
    currency: 'EUR'
  });
</script>

curl -X POST "https://api.basketbooster.eu/v1/orders" \
  -H "X-API-Key: YOUR_SECRET_KEY" -H "Content-Type: application/json" \
  -d '{
    "order_id": "ORDER_1234",
    "user_id": "cust_9",
    "session_id": "s_...",
    "currency": "EUR",
    "items": [{"item_id": "123", "quantity": 2, "price": 19.9}]
  }'

Keep the shopper id consistent

The Impact dashboard credits an add-to-cart or a sale to the recommender only when the commerce event and an earlier recommendation click or impression carry the same shopper identity and product id, within a 30-day window. Identity is user_id when present, otherwise the widget’s automatic session id. When both are sent, user_id wins.

For logged-in shoppers, pass the customer id everywhere: data-user-id on every widget embed, user_id on your add-to-cart and buy calls, and user_id on POST /v1/orders. If the widgets run anonymous while your orders carry a customer id, nothing matches and attribution stays at zero.

For anonymous shoppers the widget keeps its session id in localStorage["pr_sid"], and browser-side ProductRecs.track calls pick it up automatically. For server-side order sync, mirror the id into a first-party cookie and forward it as session_id:

<script>
  try {
    var sid = localStorage.getItem('pr_sid');
    if (sid) document.cookie = 'pr_sid=' + encodeURIComponent(sid) +
      '; path=/; max-age=15552000; SameSite=Lax';
  } catch (e) {}
</script>

Then read the cookie on your server (PHP shown here):

$sid = $_COOKIE['pr_sid'] ?? '';
if (preg_match('/^s_[A-Za-z0-9]+$/', $sid)) {
    $payload['session_id'] = $sid;
}