Target.com Find Product

Purpose

Given a free-text product query (e.g. "AirPods Pro 2", "tide pods", "lego star wars millennium falcon"), return the top matching Target.com product(s) with structured fields: title, brand, price, original (struck-through) price, average rating + rating count, TCIN (Target's internal product ID), canonical product URL (/p/.../-/A-<tcin>), primary image URL, and signals like spell correction or zero-results fallback. Read-only — never adds to cart, never submits an order.

When to Use

• "Is X on Target.com? Price, link, image."
• Daily / hourly price-check on a watch-list of Target products.
• Bulk extraction of Target's top-N results for a list of keywords.
• Comparison-shopping agents that need a Target price point alongside other retailers.
• Anywhere you would otherwise drive target.com/s?searchTerm=... in a Verified browser — the redsky JSON API returns identical data in one ~1s HTTP request without anti-bot.

Workflow

Target's web search page is a Next.js SSR shell with a JS-rendered product grid; the actual data comes from a public JSON aggregation API at redsky.target.com that the browser polls on page load. The endpoint is keyed but the key is a static public token (ff457966e64d5e877fdbad070f276d18ecec4a01) embedded in Target's JS bundles, accepts unauthenticated requests, has no per-IP captcha or PerimeterX challenge, and works equally well with or without a residential proxy. Hit it directly — the browser path is a ~100× cost premium that pays for a Health Data Consent modal you have to dismiss before the grid even paints.

1. Build the request

GET https://redsky.target.com/redsky_aggregations/v1/web/plp_search_v2
    ?key=ff457966e64d5e877fdbad070f276d18ecec4a01
    &channel=WEB
    &keyword=<URL-encoded query>
    &page=%2Fs%2F<URL-encoded query>
    &visitor_id=<any non-empty string, e.g. "skill-runner">
    &pricing_store_id=2885
    &default_purchasability_filter=true
    &count=24
    &offset=0
    &platform=desktop

Required params (server returns HTTP 400 if any are missing): key, channel, keyword, page, visitor_id, pricing_store_id. Optional but useful: count (default 24), offset (item-level, not page-level — increment by count to paginate), default_purchasability_filter=true (hides out-of-stock), sort_by (see table below), category (scope to a category id), platform=desktop|mobile.

The page param is required and must look like a real PLP path (/s/<term>); the server validates only that it starts with /s/ — anything after is decorative. visitor_id is required but any non-empty string is accepted; the server doesn't actually validate it against a known device. pricing_store_id=2885 is the Target HQ store (Minneapolis) and yields nationwide pricing — use it as a stable default unless you specifically need store-local pricing.

2. Fire the request

URL="https://redsky.target.com/redsky_aggregations/v1/web/plp_search_v2?key=ff457966e64d5e877fdbad070f276d18ecec4a01&channel=WEB&keyword=AirPods+Pro+2&page=%2Fs%2FAirPods+Pro+2&visitor_id=skill-runner&pricing_store_id=2885&default_purchasability_filter=true&count=24"

# Either path works — pick whatever the agent's runtime already has:
browse cloud fetch "$URL"              # Browserbase Fetch API, no proxy needed
# or
curl -fsS -H 'Accept: application/json' "$URL"

A residential proxy is not required for this endpoint. Verified: same response shape, same products, same status 200, with and without --proxies. Save the proxy budget for the browser-fallback path.

3. Detect outcome shape before extracting

.data.search.search_response |
  {
    total_results: .metadata.total_results,
    spell_corrected: (.metadata.auto_corrected_keyword // null),
    real_results: ((.facet_list // []) | length > 0),
    keyword: .metadata.keyword
  }

Branch on real_results:

• real_results: true — Target found genuine matches for the keyword. Parse data.search.products[].
• real_results: false — Target found zero matches and is serving generic "recommendation" filler (typically ~200 unrelated bestsellers from books / toys / random categories). Do NOT silently return the first filler item as the match. Emit success: false, reason: "no_results" and pass through metadata.keyword so the caller can decide what to do.
• spell_corrected non-null — Target auto-corrected the query (e.g. "airpoods" → "airpods"). The returned products correspond to the corrected keyword. Pass the corrected term back so the caller can decide whether the correction is acceptable.

4. Extract the top product

.data.search.products[0] |
  {
    tcin: .tcin,                                                       # e.g. "85978609"
    title: (.item.product_description.title | @html_decode),           # decode  , &, ", ™
    brand: .item.primary_brand.name,                                   # may be null
    price: .price.formatted_current_price,                             # "$249.99" or "See price in cart"
    original_price: .price.formatted_comparison_price,                 # null if not on sale
    rating: .ratings_and_reviews.statistics.rating.average,            # 0–5 float, null if no ratings
    rating_count: .ratings_and_reviews.statistics.rating.count,        # integer, null if no ratings
    product_url: .item.enrichment.buy_url,                             # canonical https://www.target.com/p/.../-/A-<tcin>
    image_url: .item.enrichment.image_info.primary_image.url,          # scene7 CDN URL
    desirability: (.desirability_cues // [] | map(.display))           # e.g. ["Bestseller"], ["5k+ bought in last month"]
  }

Decode HTML entities (  → non-breaking space, & → &, " → ", ™ → ™) — Target's API returns titles with entity references intact, not pre-decoded text.

5. (Optional) Filter sponsored

plp_search_v2 already filters most sponsored placements server-side — the field item.is_sponsored is null (not false) across all observed organic results, and sponsored slots come from a sibling aggregation (cf_sponsored_products_search_v1) the browser fetches separately. If you need to be defensive, drop any product whose __typename !== "ProductSummary".

6. Return JSON; do not navigate

Read-only is the rule. Do not hit the add_to_cart redsky aggregation. Do not POST to /checkout. Stop at the product summary.

Browser fallback

If redsky.target.com ever returns 4xx (key rotation, geo-block, etc.) — drive the public site:

1. Create a Browserbase session with --proxies --verified (PerimeterX / HUMAN sensor at client.px-cloud.net/PXGWPp4wUS/main.min.js will challenge a bare session within a few requests).
2. browse open "https://www.target.com/s?searchTerm=<URL-encoded query>" --remote.
3. Dismiss the Health Data Consent modal. On a first-visit session, Target overlays a "Health Data Consent" modal (Virginia VCDPA compliance) that blocks the product grid from being interactable. Click the "Continue shopping" button — selector button:has-text("Continue shopping") — before snapshotting.
4. browse wait timeout 2500 — the grid lazy-loads after consent dismissal.
5. browse snapshot — product cards surface as link refs with TCIN-bearing hrefs.
6. Extract the same fields by parsing the rendered DOM. The product card structure is [data-test="@web/site-top-of-funnel/ProductCardWrapper"] → child [data-test="product-title"] (title), [data-test="product-price"] (price), [data-test="ratings"] (rating).

The browser fallback costs ~6–8 turns and ~30–60s wall vs the API's one HTTP request. Use only when the API path is confirmed broken — not as the default.

Site-Specific Gotchas

• The redsky API key is public and stable. ff457966e64d5e877fdbad070f276d18ecec4a01 has been observed in Target's web bundle continuously since at least 2021 across third-party scrapers. If it ever rotates, scan the Target SERP HTML for redsky_aggregations references and look for the key in the surrounding JS — historically it's been bundled in a global config object emitted into the SSR HTML, not behind any auth wall.
• Required-param surface is fragile. Dropping any one of key, channel, keyword, page, visitor_id, pricing_store_id flips the response to HTTP 400 with no error body. Removing default_purchasability_filter, count, offset, platform, store_ids, or sort_by is safe — those default cleanly server-side.
• page is path-prefix-validated, not echo'd. The server accepts any page starting with /s/. It does not compare page against keyword. You can pass page=%2Fs%2Ftest for every request regardless of the actual keyword and the response is identical to passing the real path — verified across tide+pods, airpods+pro+2, instant+pot+duo+7+in+1.
• "No results" returns padding, not an error. A nonsense keyword (xkcd12345nonexistent) returns HTTP 200 with total_results: 200 and 24 unrelated bestsellers in products[]. The reliable signal is search_response.facet_list is missing/empty for the padding case but populated for genuine results. total_results is not a reliable zero-results signal — Target pads it.
• Spell correction is server-side and silent in products[]. When metadata.auto_corrected_keyword is non-null, the returned products are for the corrected keyword, not the typed one. The caller should be told about the correction so they can decide whether to accept it. (Example: keyword=airpoods returns auto_corrected_keyword="airpods" plus 481 AirPod-family products.)
• Titles contain raw HTML entities, not decoded text.   (nbsp), & (&), " ("), ™ (™) appear verbatim. Always run titles through an HTML-entity decoder before emitting JSON to a consumer.
• price.formatted_current_price can be "See price in cart" for MAP-restricted items (Apple, Instant Pot RIO line, some appliances). This is a real product, just legally price-suppressed on the PLP — open the PDP via buy_url to see the actual cart price, or read price.current_retail (numeric, sometimes populated even when the formatted string is suppressed).
• brand is .item.primary_brand.name, but it can be null — observed null for Target-private-label staples (e.g. "up & up"), unbranded commodity items, and grocery. Fall back to extracting the leading token of the title if a brand is required downstream.
• rating.count is the rating count, not the review count. A separate review_count field at the top of statistics does not exist on plp_search_v2 — the field .ratings_and_reviews.statistics.rating.count is the integer you want.
• pricing_store_id controls store-localized pricing, not nationwide-vs-store availability. 2885 (Jersey City / NJ) is the value Target's public web bundles use by default. Other valid store IDs (1000 = Minneapolis 50th & France, 3991 = LA Westwood, etc.) are accepted. The default_purchasability_filter parameter independently controls whether out-of-stock items are hidden.
• Health Data Consent modal blocks the browser-fallback path. Target overlays a Virginia VCDPA compliance modal on every cold-session SERP / PDP that intercepts pointer events on the entire grid. Must be dismissed via "Continue shopping" before any product card is interactable. The API path bypasses this entirely.
• Anti-bot on the public site is PerimeterX / HUMAN. client.px-cloud.net/PXGWPp4wUS/main.min.js (the HUMAN sensor) is loaded as <script id="humanSensor"> on every page. Bare sessions get challenged within ~5–10 requests; --proxies --verified is mandatory for the browser fallback. The redsky API endpoint is not behind PerimeterX — verified: 200 OK on direct curl from a datacenter IP, no challenge, no _px* cookies required.
• The PDP URL is canonical and idempotent. https://www.target.com/p/<slug>/-/A-<tcin> is stable — once you have the TCIN you can drop the slug entirely (/-/A-<tcin>) and Target 301s to the canonical slug.
• Pagination uses offset not page. offset is item-level (0, 24, 48, ...). The redsky endpoint also accepts &page_number=2 but it's a no-op — only offset controls pagination.
• store_ids filters availability, not pricing. Including store_ids=<id> scopes the result set to items in-stock at that store. Omitting it returns everything Target sells online for the keyword. They're independent dimensions from pricing_store_id.

Expected Output

Three distinct outcome shapes:

// Success — top result is a genuine match
{
  "success": true,
  "query": "AirPods Pro 2",
  "search_keyword": "AirPods Pro 2",
  "spell_corrected_from": null,
  "total_results": 259,
  "product": {
    "tcin": "85978609",
    "title": "Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation",
    "brand": "Apple",
    "price": "$249.99",
    "original_price": null,
    "rating": 3.9,
    "rating_count": 1441,
    "product_url": "https://www.target.com/p/ap2022-true-wireless-bluetooth-headphones/-/A-85978609",
    "image_url": "https://target.scene7.com/is/image/Target/GUEST_d1b8c229-751b-430b-a0fb-521d7777a784",
    "desirability_cues": ["Bestseller"],
    "is_sponsored": false
  },
  "error_reasoning": null
}

// Success with spell correction — caller should decide whether to accept
{
  "success": true,
  "query": "airpoods",
  "search_keyword": "airpods",
  "spell_corrected_from": "airpoods",
  "total_results": 481,
  "product": { /* ... top AirPods product ... */ },
  "error_reasoning": null
}

// No real matches — Target returned recommendation filler
{
  "success": false,
  "query": "xkcd12345nonexistent",
  "search_keyword": "xkcd12345nonexistent",
  "spell_corrected_from": null,
  "total_results": 200,
  "product": null,
  "error_reasoning": "no_results",
  "note": "Target returned 200 unrelated recommendations (no facet_list in search_response — confirmed filler)."
}

// API blocked — fall back to browser path (rare)
{
  "success": false,
  "query": "...",
  "product": null,
  "error_reasoning": "api_blocked",
  "http_status": 403,
  "note": "redsky.target.com returned 403. Switch to the browser-fallback workflow (Health Data Consent modal + Verified session)."
}