search-products▌
zepto.com/search-products-wtqzgb · updated May 21, 2026
MDX-style export adds YAML metadata + attribution linking explainx.ai and this canonical listing URL.
Search Zepto for products matching a query at a given Indian delivery location, returning in-stock items with name, brand, pack size, price, MRP, discount, rating, review count, stock state, ETA, product URL, and image URL. Read-only.
| name | search-products |
| title | Zepto Product Search |
| description | >- Search Zepto for products matching a query at a given Indian delivery location, returning in-stock items with name, brand, pack size, price, MRP, discount, rating, review count, stock state, ETA, product URL, and image URL. Read-only. |
| website | zepto.com |
| category | quick-commerce |
| tags | - grocery - india - quick-commerce - search - read-only |
| source | 'browserbase: agent-runtime 2026-05-19' |
| updated | '2026-05-19' |
| recommended_method | browser |
| alternative_methods | [] |
| verified | true |
| proxies | true |
Zepto Product Search
Purpose
Given a free-text product query and an Indian delivery location (street/area/landmark name or pincode), return the best-matching in-stock products listed on Zepto for that location's delivery store — with product name, brand, pack size/variant, current price, MRP, discount, rating, review count, in-stock status, per-card delivery ETA, canonical product URL, and image URL. Read-only — never clicks ADD, never proceeds toward cart/checkout.
When to Use
- "What's the price of Amul butter / Maggi noodles / Cadbury Dairy Milk on Zepto in Koramangala?"
- "Find the cheapest 1 L milk available for delivery in Bandra right now."
- Price-monitoring scrapes that need MRP + discount + per-store availability.
- Anywhere you need to know whether Zepto actually services a given Indian address before quoting a delivery ETA.
Workflow
The Zepto web site at www.zepto.com is a Next.js SPA. The recommended path is browser-driven via a Browserbase remote session, navigating directly to https://www.zepto.com/search?query=<URL-encoded query> and parsing product cards from browse snapshot. A backing JSON API exists (POST https://bff-gateway.zepto.com/user-search-service/api/v3/search) but is gated by per-session secrets — see Site-Specific Gotchas; do not try to call it cold.
-
Create the session. Use a Browserbase session with stealth + proxies, region
ap-southeast-1(Singapore — closest geographically; lowest latency to Zepto's India CDN and least likely to hit IP-mismatch heuristics). Verified + proxies is required: Zepto serves a CloudFront-fronted Next.js app that fingerprints aggressively, and the address-autocomplete + map widgets fail without a residential India-routable IP.sid=$(browse cloud sessions create --keep-alive --proxies --verified \ --region ap-southeast-1 \ | node -e "let s='';process.stdin.on('data',c=>s+=c).on('end',()=>process.stdout.write(JSON.parse(s).id))") export BROWSE_SESSION="$sid" -
Set the delivery location BEFORE searching. Zepto's IP-based default is unreliable (it picks an arbitrary metro and may resolve to a non-serviceable area), and price/availability/ETA all vary by
storeId. The site uses Google-Places-style address autocomplete, not pincode entry — you type a free-text address and pick a suggestion.browse open "https://www.zepto.com/" --remote browse wait load --remote && browse wait timeout 2000 --remote browse snapshot --remote # find ref for `button: Select Location` browse click "[<select-location-ref>]" --remote browse wait timeout 1500 --remote browse snapshot --remote # find ref for `textbox: Search a new address` browse click "[<textbox-ref>]" --remote browse type "Koramangala Bangalore" --remote browse wait timeout 2500 --remote browse snapshot --remote # 5 suggestions appear under the textbox # Click the first matching suggestion's container div ref. Suggestions look like: # [N] div # [N+1] StaticText: Koramangala # [N+2] StaticText: Koramangala, Bangalore, Karnataka, India browse click "[<first-suggestion-ref>]" --remote browse wait timeout 3000 --remoteAfter the suggestion is clicked, the modal closes, the page reloads, and the header switches from
button: Select Locationtobutton: <Area Short Name>plus aheading: <N> minutesETA badge. This is the success signal — do not search until you see that header heading. -
Search via URL. The query is a URL param; no need to click into the search box.
QUERY=$(node -e "process.stdout.write(encodeURIComponent('basmati rice'))") browse open "https://www.zepto.com/search?query=$QUERY" --remote browse wait load --remote && browse wait timeout 2500 --remote browse snapshot --remote > /tmp/zsnap.jsonThe location set in step 2 persists across the navigation — the new page header still shows the same area button and the same global ETA.
-
Branch on what the page shows. Inspect the snapshot tree for one of five top-level patterns:
Pattern in treetextOutcome heading: Showing results for "<q>"followed by productlink:rowssuccess — parse results (step 5) heading: Could not find any products for "<q>"no_results — return empty array heading: Sit Tight! We're Coming Soon!+Our team is working tirelessly…not_serviceable — Zepto does not deliver to the selected location heading: Try Again in <N> Mins(no products)store_offline — local dark store currently closed; retry later heading: Try Again in <N> Mins+ product rowsstore_offline_but_browsable — results still rendered for browsing but ordering is paused; rating/price valid, ETA NOT shown per-card -
Parse each product card. Every result is an
[X-Y] link: <inline summary>whosehref(inurlMap) points tohttps://www.zepto.com/pn/<slug>/pvid/<uuid>. The link's accessible name is a single concatenated string in this fixed order — extract via the nestedStaticTextchildren, NOT by regexing the link name (the accessible name omits some fields and concatenates without separators):image: <Product Name>— accessible alt = product name. The actual image URL is inurlMap[<image-ref>]and looks likehttps://cdn.zeptonow.com/production/ik-seo/tr:w-403,ar-...,pr-true,f-auto,q-40,dpr-2/cms/product_variant/<uuid>/<slug>.jpeg.button: ADD— present iff the product is currently in stock. An out-of-stock card replaces this with aNotify Mebutton or omits the action entirely. Always check for the literalbutton: ADDtext on the card to decideinStock.- OPTIONAL
image: P3 - Ad.png(orP1 - …,P2 - …) — sponsored/Premium ad slot. Treat the card normally but setsponsored: true. span→StaticText: ₹+StaticText: <integer>— discounted (current) selling price in ₹ (whole rupees, no decimals on the site).- OPTIONAL second
span→StaticText: ₹+StaticText: <integer>— MRP. Present only when there is a discount. If absent, treatmrp = sellingPriceanddiscount = 0. - OPTIONAL
StaticText: ₹<N>+StaticText: OFF— flat-amount discount badge. The percent can be derived as(N / mrp) * 100. StaticText: <Product Name>— canonical product name (use this, not theimage:alt —image:is sometimes truncated).StaticText: <pack-size>— e.g.1 pc (20.2 g),1 pack (500 ml),1 L,1 pack (5 kg). Free-form string; preserve verbatim.- OPTIONAL
StaticText: <variant>— sub-variant label likeMilk Chocolate,Long Grain,Silk Oreo,Fruit & Nut,Premium. Optional; may be absent on basics. - OPTIONAL rating block:
span→image(star icon) +StaticText: <rating>(e.g.4.6,5) followed byspan→StaticText: (+StaticText: <count>+StaticText: ). Review count uses K/M suffix (e.g.274.5k,1.9k,855,125.4k). Parse withparseFloat(num) * (suffix === 'k' ? 1000 : suffix === 'm' ? 1e6 : 1). Many cards have no rating block at all — treat asnull. - OPTIONAL
StaticText: <N> mins— per-card delivery ETA, e.g.9 mins. Present only when the selected store is online (i.e. the global header reads<N> minutes, notTry Again in <N> Mins). - OPTIONAL trailing
StaticText: Bestseller/StaticText: New— quality/recency tags.
Brand is not a separate field on the card. Zepto inlines the brand into the product name (e.g.
Cadbury Dairy Milk Chocolate Bar Cricket Pack,Nandini Goodlife Toned UHT Milk,Nestle Kitkat 4 Fingers …). Extract by matching the first word(s) of the product name against a known-brand list, OR — preferred for accuracy — open the product detail page (PDP) at the card's/pn/<slug>/pvid/<uuid>URL; the PDP returns the brand as a structured field. Only do PDP fetches for the top-N results the caller actually cares about; doing it for every card on a 263-result query is wasteful. -
Pagination. Results render as a vertical-feed grid of
PRODUCT_GRIDwidgets (3 cards per row, ~30 cards per page). To fetch more, scroll the page (browse mouse scroll 600 400 0 1500, wait 1500ms, re-snapshot, dedupe bypvidUUID). The site lazily loads more grids — there's no "Next page" button. Sponsored ad cards are interleaved between organic grids; tag them withsponsored: true(presence ofP3 - Ad.png/Sponsoredlabel) and exclude or rank lower depending on caller intent. -
Release the session.
browse cloud sessions update "$sid" --status REQUEST_RELEASE
API path (DO NOT use cold)
Zepto's web app calls POST https://bff-gateway.zepto.com/user-search-service/api/v3/search with body {"query":"<q>","pageNumber":0,"mode":"SHOW_ALL_RESULTS","userSessionId":"<uuid>"}. The response is a clean JSON layout with productResponse blocks containing product.brand, productVariant.formattedPacksize, productVariant.ratingSummary.{averageRating,totalRatings}, discountedSellingPrice (paise — divide by 100 to get ₹), mrp (paise), outOfStock, etc. — far richer than the rendered card. However, the request requires ~30 headers including x-csrf-secret, x-xsrf-token, request-signature (SHA hash of body + secret), storeId/store_ids (derived from the selected delivery location), deviceId, sessionId, and a 1.5KB compatible_components feature-flag string. The signature is rotated per request and validated server-side. We attempted to replay a captured request out-of-band; verified blocked. The only viable use of the API is in-page via browse eval after the same session has set its location, which is strictly more expensive than parsing the rendered DOM. Don't waste time on cold API calls.
Site-Specific Gotchas
- Location must be set BEFORE searching, not after. Without an explicit location, Zepto picks a metro from the request IP. Through a residential proxy this typically resolves to a sane Indian metro, but the specific dark-store assignment is non-deterministic across runs — price, availability, and ETA all vary. Always run the address-autocomplete flow (step 2) first and verify the header reads
<N> minutesbefore issuing search queries. - No pincode field — the picker is Google-Places autocomplete. Don't try to enter a 6-digit pincode directly; it won't typeahead. Type the area name + city (e.g.
Bandra West Mumbai,Koramangala Bangalore,Sector 18 Noida) and click the first suggestion. The textbox accepts free-form English; Hindi script also works but English is the safer canonicalization. - The Select Location button click does NOT visibly open a modal in the snapshot's first 200 lines — the modal renders far down the tree (search
treetext for the literal stringYour Location). When verifying the modal opened, look fortextbox: Search a new address, not for a top-of-tree dialog. The xpathMap puts the modal insidebody/div[2]/div[2]/...while the rest of the page lives inbody/div[2]/div[1]/.... - Picking a suggestion is a single click — no Confirm button. Some quick-commerce sites require a follow-up "Confirm address" tap. Zepto does not. The modal closes on first click of any suggestion-row container div.
- Five distinct page states for the same
/search?query=URL. Always branch on the top-levelheading:text before parsing cards.Showing results for "<q>",Could not find any products for "<q>",Sit Tight! We're Coming Soon!,Try Again in <N> Mins(alone), andTry Again in <N> Mins+ products — each requires different handling. Defaulting to "parse alllink:rows under banner" without checking the heading yields silent garbage on the not-serviceable and offline-store branches. Try Again in <N> Minsis per-store, not global. A given location's dark store can be offline temporarily (curfew hours, restocking, weather). Other locations the same minute may show9 minutesfine. IfTry Again in Nshows with no product rows, the only fix is changing location or waiting; retrying immediately will not help.- Brand is not a discrete field on cards. The accessibility tree exposes a flat product-name string. The first-word brand heuristic works for ~80% of cases (
Cadbury,Nestle,Amul,Nandini,Daawat,Hocco,Tata) but breaks for multi-word brands (Mother Dairy,Pure Temptation,Bb Royal,B Natural). For accurate brand extraction, hit the PDP at/pn/<slug>/pvid/<uuid>— the PDP page title format isBuy <Product Name> Online - Price @ ₹<N> | Instant Delivery | Zeptoand the breadcrumb under the buy box givesHome > <L2 Category> > <L3 Category>. The cleanest brand source is the API'sproductResponse.product.brandfield; if you can extract it from an in-page XHR you've already paid for, do that instead of a second PDP roundtrip. - Prices on the rendered card are whole rupees with no decimals. The backing API uses paise (
discountedSellingPrice: 1800= ₹18.00). Don't confuse the two: if you scrape the card, you have ₹; if you scrape an XHR response, divide by 100. - Rating block is often absent. Roughly half of cards lack a star + review-count block (new SKUs, low-volume categories). Treat missing as
null, not 0. P3 - Ad.png/P1 - …/P2 - …marks sponsored cards. They are real products but are paid placements. Setsponsored: trueand consider de-ranking them when computing "best match".- Review counts use Indian-style K suffix, not commas.
274.5k= 274,500.1.9k= 1,900.42.3k= 42,300. Parse before storing. - Image URLs in
urlMapare ImageKit-transformed CDN paths. They include resize/quality params (tr:w-403,ar-4000-4000,pr-true,f-auto,q-40,dpr-2). If you want a higher-res version, strip thetr:…/segment to get the original path undercms/product_variant/<uuid>/<slug>.jpegatcdn.zeptonow.com/production/. - The backing API at
bff-gateway.zepto.com/user-search-service/api/v3/searchis signed and gated. Headers include a rotatingx-csrf-secret, a body-signedrequest-signature, store IDs derived from the selected delivery location, and a 1.5KBcompatible_componentsflag set. Cold curl/fetch returns 401/403 (or worse — a 200 with empty/sanitised payload). Don't bother trying to spoof; either parse the rendered DOM, or usebrowse evalfrom within an already-authenticated session. bff-gateway.zepto.comandapi.zeptonow.comare NOT the same surface. The latter returns 500 on every cold path we probed; onlybff-gatewaycarries the v3 search API, and only behind in-page auth.- CSP locks scripts to
*.zeptonow.com,*.zeptonow.dev,*.zepto.com,cdn.zeptonow.com. Don't try to inject third-party JS for scraping;browse eval(page-context) works because it runs as the page's own origin, but anything else will get CSP-blocked. - Proxy region matters.
--region ap-southeast-1(Singapore) is closest to Zepto's India CloudFront edges and the address-suggestion latency is bearable (~2.5s for the 5-item dropdown). EU/US regions add 600-900ms of latency on every modal interaction and sometimes time out the autocomplete typeahead. - The page lazily renders cards as you scroll — a fresh
/search?query=snapshot returns only the first ~30 cards (10PRODUCT_GRIDwidgets of 3 each). To get the full 263-result set for a popular query like "chocolate", you must scroll-paginate; there is no?page=NURL param that works. - Cookies/session state persist across
browse opencalls within the same session. Setting location once means subsequent/search?query=…navigations keep that location. This is the only reason the URL-search step is cheap; if the session restarts, redo step 2. - READ-ONLY. Never click
ADD, never proceed to/cartor/checkout. Cart-modification XHRs are gated by the same signed-header machinery as search and don't easily round-trip back.
Expected Output
Five distinct outcome shapes; the caller should branch on status:
// success — normal results
{
"status": "success",
"query": "basmati rice",
"location": {
"label": "Koramangala",
"fullAddress": "Koramangala, Bangalore, Karnataka",
"etaMinutes": 9
},
"totalShown": 30,
"totalAvailable": 263,
"products": [
{
"name": "Daawat Pulav Basmati Rice | Long Grain",
"brand": "Daawat",
"packSize": "1 pack (500 g)",
"variant": "Long Grain",
"price": 75,
"mrp": 90,
"discount": { "amount": 15, "percent": 17 },
"rating": 4.6,
"reviewCount": 16800,
"inStock": true,
"etaMinutes": 9,
"sponsored": false,
"tags": [],
"url": "https://www.zepto.com/pn/daawat-pulav-basmati-rice-long-grain/pvid/96b12acf-bdcd-47bb-a9e9-a04364fd30e1",
"imageUrl": "https://cdn.zeptonow.com/production/ik-seo/tr:w-403,ar-3000-3000,pr-true,f-auto,q-40,dpr-2/cms/product_variant/19b0b091-07e3-40b6-b60e-0719e39caf64/Daawat-Pulav-Basmati-Rice-Long-Grain.jpeg"
}
]
}
// no_results — query matched zero SKUs in this location
{
"status": "no_results",
"query": "xyzzyplugh",
"location": { "label": "Bandra C", "fullAddress": "Bandra C, Mumbai, Maharashtra", "etaMinutes": 9 },
"products": []
}
// not_serviceable — Zepto doesn't deliver to the chosen address
{
"status": "not_serviceable",
"query": "milk",
"location": { "label": "Other", "fullAddress": "Andaman and Nicobar Islands", "etaMinutes": null },
"message": "Sit Tight! We're Coming Soon!"
}
// store_offline — local dark store temporarily closed, no products shown
{
"status": "store_offline",
"query": "coffee",
"location": { "label": "Koramangala", "fullAddress": "Koramangala, Bangalore, Karnataka", "etaMinutes": null },
"retryInMinutes": 15,
"products": []
}
// store_offline_but_browsable — store paused but the catalog still renders
// for browsing. Cards have valid price/MRP/rating but NO per-card ETA;
// inStock should be reported as `null` (intent: orderable when reopened),
// not `true` — placing an order would be blocked at checkout time.
{
"status": "store_offline_but_browsable",
"query": "coffee",
"location": { "label": "Koramangala", "fullAddress": "Koramangala, Bangalore, Karnataka", "etaMinutes": null },
"retryInMinutes": 15,
"products": [
{
"name": "Nescafe Classic - Instant Coffee Powder - Pure Coffee",
"brand": "Nescafe",
"packSize": "1 pc (24 g)",
"variant": null,
"price": 116,
"mrp": 124,
"discount": { "amount": 8, "percent": 6 },
"rating": 4.7,
"reviewCount": 12800,
"inStock": null,
"etaMinutes": null,
"sponsored": false,
"tags": [],
"url": "https://www.zepto.com/pn/nescafe-classic-instant-coffee-powder/pvid/<uuid>",
"imageUrl": "https://cdn.zeptonow.com/production/ik-seo/tr:w-403,ar-1200-1200,pr-true,f-auto,q-40,dpr-2/cms/product_variant/<uuid>/Nescafe-Classic.jpeg"
}
]
}
How to use search-products on Cursor
AI-first code editor with Composer
Prerequisites
Before installing skills in Cursor, ensure your development environment meets these requirements:
- ›Cursor installed and configured on your development machine
- ›Node.js version 16.0+ with npm package manager (verify with
node --version) - ›Active project directory or workspace where you want to add search-products
Execute installation command
Execute the skills CLI command in your project's root directory to begin installation:
The skills CLI fetches search-products from GitHub repository zepto.com/search-products-wtqzgb and configures it for Cursor.
Select Cursor when prompted
The CLI will show a list of available agents. Use arrow keys to navigate and space to select Cursor:
Verify installation
Confirm successful installation by checking the skill directory location:
Reload or restart Cursor to activate search-products. Access the skill through slash commands (e.g., /search-products) or your agent's skill management interface.
Security & Verification Notice
We perform automated surface-level scans (Gen AI Scanner, Socket, Snyk) during installation. These checks detect common vulnerabilities but do not guarantee complete security. Always review skill source code and verify the publisher's reputation before production use.
Skills execute code in your development environment. Always verify the publisher's identity, review recent commits, and test in isolated environments before production deployment.
List & Monetize Your Skill
Submit your Claude Code skill and start earning
Use Cases▌
Task Automation & Efficiency
Automate repetitive workflows and reduce manual effort
Example
Generate reports, summarize documents, draft communications
Save 3-5 hours per week on routine tasks
Knowledge Enhancement
Learn new skills, understand complex topics, get expert guidance
Example
Explain concepts, provide examples, suggest learning resources
Accelerate learning and skill development by 2x
Quality Improvement
Enhance output quality through reviews, suggestions, and refinements
Example
Review drafts, suggest improvements, catch errors
Improve work quality by 30-40% with less effort
Implementation Guide▌
Prerequisites
- ›Claude Desktop or compatible AI client with skill support
- ›Clear understanding of task or problem to solve
- ›Willingness to iterate and refine outputs
Time Estimate
15-45 minutes depending on use case complexity
Installation Steps
- 1.Install skill using provided installation command
- 2.Test with simple use case relevant to your work
- 3.Evaluate output quality and relevance
- 4.Iterate on prompts to improve results
- 5.Integrate into regular workflow if valuable
Common Pitfalls
- ⚠Expecting perfect results without iteration
- ⚠Not providing enough context in prompts
- ⚠Using skill for tasks outside its intended scope
- ⚠Accepting outputs without review and validation
Best Practices▌
✓ Do
- +Start with clear, specific prompts
- +Provide relevant context and constraints
- +Review and refine all outputs before using
- +Iterate to improve output quality
- +Document successful prompt patterns
✗ Don't
- −Don't use without understanding skill limitations
- −Don't skip validation of outputs
- −Don't share sensitive information in prompts
- −Don't expect skill to replace human judgment
💡 Pro Tips
- ★Be specific about desired format and style
- ★Ask for multiple options to choose from
- ★Request explanations to understand reasoning
- ★Combine AI efficiency with human expertise
When to Use This▌
✓ Use When
Use when skill capabilities match your task, clear ROI on time saved, and you can validate outputs. Best for repetitive tasks, learning, and quality improvement.
✗ Avoid When
Avoid when task requires deep expertise you can't validate, involves sensitive decisions, or when learning process is more valuable than speed of completion.
Learning Path▌
- 1Familiarize yourself with skill capabilities and limitations
- 2Start with low-risk, non-critical tasks
- 3Progress to more complex and valuable use cases
- 4Build expertise through regular use and experimentation
Discussion
Product Hunt–style comments (not star reviews)- No comments yet — start the thread.
Ratings
4.5★★★★★42 reviews- ★★★★★Dhruvi Jain· Dec 24, 2024
search-products is among the better-maintained entries we tried; worth keeping pinned for repeat workflows.
- ★★★★★Yuki Srinivasan· Dec 16, 2024
Useful defaults in search-products — fewer surprises than typical one-off scripts, and it plays nicely with `npx skills` flows.
- ★★★★★Sakura Gupta· Dec 8, 2024
Solid pick for teams standardizing on skills: search-products is focused, and the summary matches what you get after install.
- ★★★★★Sophia Jain· Dec 4, 2024
Registry listing for search-products matched our evaluation — installs cleanly and behaves as described in the markdown.
- ★★★★★Yuki Singh· Nov 27, 2024
search-products has been reliable in day-to-day use. Documentation quality is above average for community skills.
- ★★★★★Nia Torres· Nov 23, 2024
search-products reduced setup friction for our internal harness; good balance of opinion and flexibility.
- ★★★★★Oshnikdeep· Nov 15, 2024
search-products fits our agent workflows well — practical, well scoped, and easy to wire into existing repos.
- ★★★★★Rahul Santra· Nov 7, 2024
Keeps context tight: search-products is the kind of skill you can hand to a new teammate without a long onboarding doc.
- ★★★★★Omar Anderson· Nov 7, 2024
We added search-products from the explainx registry; install was straightforward and the SKILL.md answered most questions upfront.
- ★★★★★Pratham Ware· Oct 26, 2024
I recommend search-products for anyone iterating fast on agent tooling; clear intent and a small, reviewable surface area.
showing 1-10 of 42