50 AI agents for Shopify: run the store jobs you never get to.

All 50 agents, mapped

The six deep dives below are full setup guides.

Catalog & product data · 7 agents

• Product title — rewrites titles to the search-first pattern: type, key attribute, brand, variant
• Description writer — benefit-led descriptions from specs and reviews, one voice across the catalog
• Collection page — collection copy and H1s that rank — no more empty collection templates
• Metafield mapper — fills the metafields your theme and feeds actually read
• Image alt text — alt text for every product image, from the product data, not guesses
• Feed health — missing GTINs, bad categories, price mismatches — before Google flags them
• Merchant center — disapprovals and warnings, grouped by cause, with the fix per group

Pricing & merchandising · 4 agents

• Price testing — designs price tests by product, reads the results, calls the winner
• Bundle builder — finds products bought together and drafts the bundle, price and page
• Upsell flow — the post-purchase and cart upsells that actually get taken
• Cross sell — the “goes with” picks per product, from real order pairs

Cart & checkout · 3 agents

• Cart recovery — reads abandoned checkouts, tells you why they left, fixes the flow
• Checkout audit — walks checkout as a customer, flags every friction point
• Shipping rules — rate tables vs order data — where shipping cost is losing you orders

Inventory & margin · 6 agents

• Inventory alert — days-of-cover per SKU, alerts before the stockout, not after
• Restock forecast — PO quantities per SKU from velocity, seasonality and lead time
• Dead stock — what hasn’t moved in 90 days, and the clearance plan for it
• Margin guard — true margin per order after discounts, shipping and fees — leaks named
• Discount strategy — which codes make money and which just discount buyers who’d have paid
• Promo schedule — a promo calendar that stops stacking and covers slow weeks

Retention & lifecycle · 9 agents

• Klaviyo flow — audits welcome, browse, cart and post-purchase flows against your data
• Email segment — segments from order behavior, not list tags
• SMS campaign — short campaigns tied to restocks, drops and last-chance windows
• Review request — asks the right customers at the right time — after delivery, not before
• UGC collector — gathers customer photos and quotes into usable product-page assets
• Loyalty program — tiers and rewards from real repeat behavior, not templates
• Subscription churn — who’s about to cancel and the save offer for each cohort
• LTV cohort — lifetime value by acquisition month and channel
• RFM segment — recency, frequency, monetary — the customer list sorted into who to email

Storefront & CRO · 8 agents

• CRO audit — the store walk-through: 40 checks, ranked by expected lift
• PDP optimizer — product pages scored and fixed — above the fold, proof, shipping, FAQ
• AB test — designs the test, sizes the sample, reads significance honestly
• Landing page — campaign landing pages that match the ad, from your own products
• Site speed — what’s slow, why, and which apps and images are the cause
• Theme QA — checks every template after a theme update — broken sections, missing blocks
• App bloat — apps you pay for, don’t use, and that slow the store down
• Search filter — on-site search terms with zero results, and the products to map them to

Support & risk · 6 agents

• Size guide — size guides from returns data — the sizes people actually swap
• FAQ builder — FAQ blocks from real support tickets, per product and per policy
• Support macro — reply templates for the ten questions that fill your inbox
• Return reason — return reasons clustered by product — what to fix on the page or the product
• Fraud screen — high-risk orders flagged with the reason, before you ship
• Chargeback — evidence packs assembled from order, shipping and message data

Growth channels · 5 agents

• Store SEO — titles, metas, canonicals, collections — the Shopify SEO checklist, run weekly
• Blog content — posts that answer buyer questions and link to the right collections
• Influencer seeding — creators to seed, from who your customers already follow
• Ads feed sync — keeps Google and Meta catalogs matching the store — price, stock, images
• Marketplace sync — listings on Amazon, TikTok Shop and Etsy kept in step with Shopify

Reporting · 2 agents

• Weekly KPI — the Monday store summary: revenue, AOV, conversion, top movers, why
• Store audit — the whole store, top to bottom, as a ranked fix queue — the one to run first

Start a store

One prompt builds, runs and markets your store

AI agents set it up, handle the orders and run the ads. You approve with one click.

A portable workspace for digital nomadsDesk gear for nomads

Start a store

• Store
• Orders
• Ads

The six agents doing the heaviest lifting

These six move the most revenue in a Shopify store. Each entry is the complete setup guide: what the agent reads, what it ships, and the full role — paste it into Claude and it’s live.

01. Store audit agent: the whole store, top to bottom, as a ranked fix queue

Most stores have never had a full pass. The audit agent walks every layer — catalog data, collections, product pages, checkout, shipping rules, apps, speed, SEO basics — and turns what it finds into one queue, ranked by revenue impact and effort. It’s the run to do first: every other agent on this page works off its findings.

• Reads Shopify products, collections, orders (90 days), theme settings, installed apps + GA4 landing pages and Search Console coverage
• Ships The fix queue: issue, where it lives, evidence, expected impact, effort — top 20 first, then the long tail by category
• Cadence Monthly in full; the top 20 re-checked weekly

You are my Shopify Store Auditor.

Job: once a month, audit the whole store and give me one ranked list of what to fix.

Process:
1. Pull products, collections, 90 days of orders, theme settings and installed apps from Shopify; landing pages from GA4; coverage from Search Console.
2. Check every layer against the checklist: catalog data (titles, descriptions, images, alt text, GTINs, metafields), collections (copy, sort order, empty or thin), product pages (above the fold, proof, shipping info, FAQ), checkout and shipping rules, apps (unused, duplicated, slow), speed, SEO basics (titles, metas, canonicals, redirects).
3. For each issue found, record: where it lives (URL or handle), the evidence, expected revenue impact (high / medium / low, with the reason), and effort (minutes / hours / days).
4. Rank by impact first, effort second.
5. Diff against last month's list: fixed, still open, new.

Output: the top 20 as a table (issue, location, evidence, impact, effort, owner), then the full list grouped by category. Keep the checklist and the ranking rule identical every run so months are comparable.

02. PDP optimizer agent: scores every product page and rewrites the ones that leak

Product pages are where the money is decided, and most of them were written once and never touched. The optimizer scores each page on the things that move conversion — the first screen, proof, shipping and returns clarity, the questions buyers ask — then rewrites the weakest ones, using your own reviews and support tickets for the copy.

• Reads Shopify products + reviews, GA4 product-page views → add-to-cart → purchase, support tickets tagged to products
• Ships A scored list of product pages (0–100, with the sub-scores), and for the bottom ten: the rewritten title, first paragraph, bullets, FAQ block and the missing proof
• Cadence Weekly — ten pages a week clears a 300-SKU catalog in a quarter

You are my PDP Optimizer.

Job: every week, find the product pages that lose the most buyers and fix them.

Process:
1. Pull product pages and their views, add-to-cart rate and purchase rate for the last 30 days. Rank by (views × the gap between the page's add-to-cart rate and the store median).
2. Score each of the top 25 on: first screen (title, price, key benefit, primary image), proof (reviews, ratings, UGC), clarity (shipping, returns, sizing, delivery time), objections (the FAQ answers the questions in support tickets), and page basics (alt text, structured data, speed).
3. For the ten worst, rewrite: the title (search-first pattern), the first paragraph (benefit, for whom, why this one), five bullets from specs and reviews, an FAQ block from real tickets, and name the proof that is missing.
4. Keep the brand voice from the pages that convert best.

Output: the scoring table for the top 25 (page, views, ATC rate, score, weakest area), then the ten rewrites, ready to paste into Shopify. Same scoring rubric every run.

03. Cart recovery agent: reads abandoned checkouts and fixes the reason, not just the email

Seven in ten carts are abandoned, and the usual answer is one more reminder email. The recovery agent reads the abandoned checkouts themselves — what was in them, where people stopped, what shipping they were shown — and tells you which cause is biggest this week, then rewrites the recovery flow around it.

• Reads Shopify abandoned checkouts (30 days), shipping rates shown, discount codes attempted, plus the current recovery flow copy and timing
• Ships Abandonment causes ranked by lost revenue (shipping cost, no express option, code failed, sizing doubt, price check), and the rewritten three-message flow with timing and the offer rule
• Cadence Weekly

You are my Cart Recovery agent.

Job: every week, tell me why carts were abandoned and what to change so fewer are.

Process:
1. Pull abandoned checkouts from the last 30 days: items, cart value, step reached, shipping options shown, discount codes attempted, device, country.
2. Cluster them by likely cause: shipping cost shock, no fast option, discount code failed, sizing or fit doubt (returns data), price comparison (came back later), payment method missing.
3. Rank causes by recovered-revenue potential (carts × value × a realistic recovery rate for that cause).
4. Review the current recovery flow: timing, copy, offer. Rewrite it around the top cause: message 1 (reminder, no discount), message 2 (the objection handled), message 3 (the offer, if the rule allows).
5. Note the on-site fixes that remove the cause entirely — a shipping threshold, a size guide, an express option.

Output: the cause table (cause, carts, value, recovery estimate), the three-message flow with send times, and the on-site fix list. Same clusters every week so I can watch them move.

Store

Create a page for the Heather restock

/heather-hoodie · drafted from 4 photos

RejectApprove

Customer success

Refund Sarah K. for a damaged hoodie

Order #1042 · $68 · photo attached

RejectApprove

Ads

Raise TikTok budget to $60 a day

ROAS 4.1 over the last 7 days

RejectApprove

Start a store

Create your store from a single prompt

Say what you want to sell. Nation writes the brand, lists the products and connects checkout.

A portable workspace for digital nomadsDesk gear for nomads

Start a store

• Brand
• Products
• Checkout

04. Restock forecast agent: PO quantities per SKU from velocity, seasonality and lead time

Stockouts on your best sellers and pallets of the slow ones — the two ways inventory quietly costs a store. The forecast agent turns order history into a purchase order you could send today: per SKU, how many, by when, with the assumptions written down so you can argue with them.

• Reads Shopify inventory levels and 12 months of orders by SKU, supplier lead times you give it once, planned promos
• Ships The restock table: SKU, on hand, daily velocity, days of cover, reorder date, quantity — plus a stockout-risk list for the next 30 days
• Cadence Weekly, and before every promo

You are my Restock Forecaster.

Job: every week, tell me what to reorder, how much, and by when.

Process:
1. Pull current inventory by SKU and 12 months of orders by SKU from Shopify.
2. For each SKU compute: 30-day and 90-day daily velocity, the seasonal factor from the same weeks last year, and days of cover at current velocity.
3. Using the lead time I gave you per supplier (and a safety stock of 2 weeks unless I say otherwise), find the reorder date: the day cover drops below lead time + safety.
4. Recommend the order quantity: enough to cover lead time + the review period at forecast velocity, adjusted for planned promos I list.
5. Flag: SKUs that will stock out in the next 30 days at any order date, and SKUs with more than 180 days of cover (over-stocked).

Output: the restock table (SKU, on hand, velocity, cover, reorder by, quantity, confidence), then the 30-day stockout-risk list and the over-stock list. State every assumption in one line at the top. Same method every run.

05. Margin guard agent: true margin per order — discounts, shipping and fees, leaks named

Revenue is easy to see; margin isn’t. Discount codes stack, free-shipping thresholds sit in the wrong place, one product line quietly costs more to ship than it earns. The guard computes contribution margin per order and finds where it’s leaking — then names the rule change that closes each leak.

• Reads Shopify orders with discounts, shipping charged vs cost, refunds; product costs from Shopify or the sheet you give it; payment and app fees
• Ships Margin by product, by discount code, by shipping zone; the leak list ranked by dollars per month; the rule change for each
• Cadence Weekly

You are my Margin Guard.

Job: every week, find where the store is losing margin and tell me the rule that stops it.

Process:
1. Pull last 30 days of orders: line items, discounts applied (code and automatic), shipping charged, refunds; product cost per SKU; payment fees; the shipping cost table I gave you.
2. Compute contribution margin per order: revenue − COGS − shipping cost − payment fees − refunds − discount.
3. Cut it three ways: by product (which lines are below target margin), by discount code (which codes are used by customers who would have paid full price — check repeat buyers), by shipping zone and threshold (where free shipping costs more than the order earns).
4. Rank every leak by dollars per month.
5. For each leak propose one rule change: raise the free-shipping threshold to X, cap code stacking, exclude a collection from a code, drop or reprice a product.

Output: the three margin tables, then the leak list (leak, $/month, rule change, side effect to watch). Same definitions every week so the trend is real.

06. Weekly KPI agent: the Monday store summary — the numbers, and why they moved

Every dashboard shows the numbers; none of them says why. The KPI agent writes the Monday summary a good ops lead would: revenue, orders, AOV, conversion, top movers — each with the cause traced back to a product, a channel, a promo or a stockout — and the three things to do this week.

• Reads Shopify orders and inventory, GA4 sessions and conversion by channel, Google and Meta spend if connected
• Ships One page: KPI table (this week, last week, same week last year), the five biggest movers with the cause, and three actions
• Cadence Every Monday, before the team meets

You are my Weekly KPI agent.

Job: every Monday, write the store summary for last week and explain what moved.

Process:
1. Pull last week, the week before, and the same week last year: revenue, orders, AOV, sessions, conversion rate, returning-customer share, and ad spend by channel if connected.
2. Find the five biggest movers (up or down) in revenue by product and by channel.
3. For each mover, trace the cause: a promo, a stockout, an ad change, a landing page change, a price change, or seasonality (compare to last year). Say which, with the evidence.
4. Check the watch-list from previous weeks: did the flagged issues resolve?
5. Write three actions for this week, each tied to a mover.

Output: the KPI table (metric, this week, last week, YoY, delta), then the movers table (mover, change, cause, evidence), then three actions. One page. Same metrics and layout every week.

BuildsPublished

Marcus Wears

Drop 03 / Garage / Live now

Sewn in
a garage

RunsShipped

Order #1042Garage hoodie · $68.00

PaidShippedDelivered

MarketsLive

MWmarcuswearsSponsored

Sewn in a garage.

Shop now

Start a store

Build your dropshipping store that runs itself

Nation finds the products, builds the store, runs the ads and ships every order.

A portable workspace for digital nomadsDesk gear for nomads

Start a store

• No stock
• No money upfront
• Supplier ships

What separates an agent from a prompt

A prompt is a one-off request. An agent is a standing role with four parts — take any prompt you already use and add the missing three:

Every setup guide in this library is those four parts pre-written. That's the whole trick — and it's why the six roles above aren't "just prompts": each one names its data source, its deliverable, and its schedule.