Browse and Paginate Farfetch Listings

Purpose

Read-only skill that walks a Farfetch category/listing page and paginates through the full result set, extracting the product cards on each page (brand, product name, price, sale price, discount, availability, product URL) plus the pagination state (current page, total pages, next-page URL). It does not add to cart, log in, or buy — it stops at the listing grid.

When to Use

Enumerating all products in a Farfetch category (e.g. women's shoes, men's bags, a brand page).
Collecting product cards page-by-page for indexing, price monitoring, or assortment analysis.
Determining how many pages / roughly how many items a category contains.
Following pagination on any …/items.aspx listing, a …/sets/….aspx edit, a brand page, or a search results page.

Workflow

The fast, reliable mechanism is URL-parameter pagination — Farfetch paginates entirely via ?page=N, and deep-linking to any page works directly (verified for ?page=2 and ?page=443). You never need to click "Next"; just navigate the URL and increment page. There is no need to reverse-engineer a private JSON API — the rendered listing page exposes everything, and the internal product API sits behind Akamai bot manager.

Start a stealth remote session. Farfetch is Akamai-protected. Create the Browserbase session with --verified --proxies. Listing pages render fine with stealth on.
Open the listing page.
```
browse open "https://www.farfetch.com/shopping/women/shoes-1/items.aspx" --remote
```
URL shape is https://www.farfetch.com/shopping/{gender}/{category}/items.aspx, where {gender} is women / men / kids and {category} is a slug like shoes-1, bags-purses-1, clothing-1, trainers-1. Brand and set pages (/shopping/women/{brand}/{category}-1/items.aspx, /sets/{slug}.aspx) paginate identically.
Lazy-load the grid. Product cards are NOT all present on first paint (~12–22 are). Trigger the lazy loader with real keyboard scroll events — press End (or PageDown) ~12–18 times with ~0.7s pauses between presses. After a full scroll, ~90 product cards are in the DOM.
- Important: in-page window.scrollBy() / scrollTo() via browse eval does NOT trigger the loader — only genuine key/scroll events do.
Read the pagination state. The footer control reads <current> of <total> (e.g. 2 of 931) with Previous / Next links. The Next href is ?page=<current+1>. Use <total> as the loop bound.

Extract the product cards. Cards are anchors whose href matches -item-<digits>.aspx. Run a single browse eval that returns JSON to avoid burning turns:

browse eval '(() => {
  const grid = [...document.querySelectorAll("a[href]")]
    .filter(a => /-item-\d+\.aspx/.test(a.getAttribute("href")||""));
  const seen = new Set(); const items = [];
  for (const a of grid) {
    const href = a.getAttribute("href");
    const id = (href.match(/-item-(\d+)/)||[])[1];
    if (!id || seen.has(id)) continue; seen.add(id);
    items.push({ id, url: new URL(href, location.origin).href,
                 lines: (a.innerText||"").split("\n").map(s=>s.trim()).filter(Boolean) });
  }
  const p = document.body.innerText.match(/(\d[\d,]*) of (\d[\d,]*)/) || [];
  return JSON.stringify({ current_page: p[1]||null, total_pages: p[2]||null,
                          count: items.length, items });
})()'

Each card's lines array is roughly [badge?, brand, name, price, salePrice?, discount?, availability, "See all sizes"] — e.g. ["NeroGiardini","80mm leather wedge sandals","$279","$210","-20%","Available","See all sizes"]. Map fields best-effort: first token may be a badge (New Season, Featured); the $… tokens are price then sale price; a -N% token is the discount.

Advance to the next page. Navigate directly to the incremented URL — do not click Next:
```
browse open "https://www.farfetch.com/shopping/women/shoes-1/items.aspx?page=2" --remote
```
Repeat steps 3–5. Continue while current_page < total_pages, or until you've collected the desired number of items.
Emit the aggregated JSON (see Expected Output). Stop at the listing grid — never proceed to a product detail purchase flow.

Site-Specific Gotchas

Pagination is ?page=N only. Deep-links to arbitrary pages work (?page=443 → title Page 443 | …). No cursor/token. Page 1 omits the param; the canonical URL for page 1 has no ?page.
Lazy load needs real scroll events. browse get markdown body and browse snapshot without scrolling only see ~12–22 of ~90 items. Scroll with keyboard End/PageDown; window.scrollBy() in eval is ignored by the loader.
Recommendations carousel contaminates anchor counts. A "Recommendations" / "You may also like" carousel at the page bottom adds ~12–18 extra -item-…aspx anchors that are NOT part of the listing grid. The main grid is the first ~90 unique IDs (they appear before the carousel in DOM order); slice/cap to the grid to exclude them. The carousel also injects its own N of 12 pagination labels — match the FIRST \d+ of \d+ (e.g. 2 of 931) for the real listing pagination, not the carousel's 1 of 12.
Per-page size ≈ 90 products. 931 pages × ~90 ≈ 84k items for women's shoes — consistent with the displayed total.
Akamai bot manager is active. The homepage probe shows "no antibot" (it 301-redirects), but …/items.aspx paths sit behind Akamai (X-Akamai-Transformed, ak_p server-timing). Run remote with --verified --proxies; pages loaded reliably with stealth on. A bare HTTP fetch of a listing URL returns 4xx/410 — don't bother fetching; drive a real browser.
Don't waste time hunting for a public API. The listing data is fully present in the rendered DOM; the internal product API is Akamai-gated. URL-param pagination + DOM extraction is the supported path.
Page titles aid verification: page 1 is Designer shoes for women | FARFETCH; page N is Page N | Designer shoes for women | FARFETCH.
Turn budget when driving with an agent: do not press End one key per turn in a long loop and then snapshot repeatedly — that exhausts a limited turn budget before extraction. Batch ~6 presses, then extract everything in ONE eval call. (Two autobrowse runs hit the 30-turn cap by over-scrolling before extracting.)

Expected Output

A JSON object per page (or an aggregated array across pages):

{
  "success": true,
  "listing_url": "https://www.farfetch.com/shopping/women/shoes-1/items.aspx",
  "current_page": 2,
  "total_pages": 931,
  "next_page_url": "https://www.farfetch.com/shopping/women/shoes-1/items.aspx?page=3",
  "items_per_page": 90,
  "items": [
    {
      "id": "34126022",
      "brand": "Saint Laurent",
      "name": "Babylone strappy sandals",
      "price": "$1,178",
      "sale_price": null,
      "discount": null,
      "badge": null,
      "availability": "Available",
      "url": "https://www.farfetch.com/shopping/women/saint-laurent-babylone-strappy-sandals-item-34126022.aspx"
    },
    {
      "id": "34089662",
      "brand": "NeroGiardini",
      "name": "80mm leather wedge sandals",
      "price": "$279",
      "sale_price": "$210",
      "discount": "-20%",
      "badge": null,
      "availability": "Available",
      "url": "https://www.farfetch.com/shopping/women/nerogiardini-80mm-leather-wedge-sandals-item-34089662.aspx"
    }
  ]
}

On the last page, current_page === total_pages and next_page_url is null.

Failure shape (e.g. blocked by anti-bot or an invalid category):

{
  "success": false,
  "listing_url": "https://www.farfetch.com/shopping/women/shoes-1/items.aspx",
  "current_page": null,
  "total_pages": null,
  "items": [],
  "error_reasoning": "Akamai challenge / 4xx returned for listing page — retry with --verified --proxies."
}