Connect to via GraphQL or MCP to build your own flexible system
The self-learningperformance marketing system
Loopcraft finds ad ideas for your brand, makes the ads and tests them with your approval. Each result teaches it what your customers respond to, so it can tell you what to make next.
Website optionalNo card to startYour first full test is free
Know what worked.Know what to test next.
Example: with the same picture, the ad that led with the taste got 900 clicks to 720 for the ad that led with the mood. Loopcraft turns that into what it knows about your customers and plans the next test for your approval.
Example for Sundays, a fictional soda brand. Test 1 showed two ads with the same picture and two messages, and ended at its $900 limit: leading with the taste got 900 clicks against 720 for leading with the mood, at 45,000 views and $450 each. Before the test, outside evidence said the taste message gets about −8% to +28% more clicks; with the test, Loopcraft now reads +11% to +31%. You confirm the reading. Loopcraft then recommends testing a new reason to buy, Sundays as a weekend drink instead of alcohol, against the taste, for $600 over 14 days: 31% of reviews say it, and no ad says it yet. You accept, and Test 2 waits in Tests for your approval of $600 before anything spends.
SSundaysExampleSearch ads, briefs or askK$900 of $1,500 in Oct1,240 creditsMO
Result to confirm1 / 6
The taste message got 25% more clicks. Confirm the reading?
Test 1 met its rule at its $900 limit on Oct 12. The taste won with one picture, so the belief is promising, not settled: Test 2 checks it again on two new pictures. Purchases went the same way, too few to judge alone.
Sundays, slow and sunny.
AMood message
Clicks
720
Views
45,000
Click rate
1.60%
Spent
$450
Purchases, illustrative
13
Real yuzu, lightly sweet.More clicks
BTaste message
Clicks
900
Views
45,000
Click rate
2.00%
Spent
$450
Purchases, illustrative
19
What Loopcraft believes now
Illustrative
Leading with the taste likely brings +11% to +31% more clicks than the mood.
+11%+20%+31%
−20%No difference+20%+40%
Before your test: outside evidence
Your test alone
Now: everything together
Range narrowed from 36 to 20 points
Why the range is narrower than your test alone
Your test counts for 74%. Three outside sources agreed, more cautiously:
Example benchmark · Flavour-led drink ads15%
Your account history · 10 past Sundays ads6%
Competitor ads · 12 long-running drink ads5%
If you confirm
“Lead with the taste” becomes promising for clicks
Customers marks the taste reason “Promising after Test 1”
Tests in Learn re-rank on the new range
Recommended test2 / 6
Test the weekend swap against the taste?
31% of your reviews call Sundays their weekend drink instead of alcohol, and none of your ads says it. If it wins, it can reach people your taste ads miss. Accepting puts Test 2 in Tests as Planned.
ATaste message· control
Real yuzu, lightly sweet.
Afternoon light
Real yuzu, lightly sweet.
Pink collage
BSwap message· new
Your weekend drink. No alcohol.
Fridge grab
Your weekend drink. No alcohol.
Window light
Two pictures per message, so the result is about the message.
Type
Concept test, 2 flexible ads, 2 pictures each
Judged by
Cost per add to cart · purchases a guardrail
Split
Meta A/B test, each person sees one side
What it would teach
Illustrative
Swap vs tasterange 53 → 44 pts
−20%No difference+20%+40%
Before your test: outside evidence
Expected after this test
What this ranking rests on
Illustrative
Your tests20%
Your reviews35%
Your history0%
Benchmarks25%
Competitors20%
Learns
1.5
per $100 · ranks 1st
Costs
$600
14 days · 16 credits for 4 drafts
Next best
0.5
per $100 · ranks 2nd
Fits October
Yes
uses the last $600 of the cap
SSundaysExampleSearch ads, briefs or askK$900 of $1,500 in Oct1,240 creditsMO
New
Do more people add to cart when we sell the weekend swap instead of the taste?
Planned· waiting for your approval of $600 · accepted in Learn just now · plan drafted by Loopcraft
FromTest 1: taste beat mood, +11% to +31%67 of 214 reviews: a weekend drink instead of alcohol4 of 9 long-running competitor ads sell a swap
ATaste message· control
Real yuzu, lightly sweet.
Afternoon light
Real yuzu, lightly sweet.
Pink collage
BSwap message· new
Your weekend drink. No alcohol.
Fridge grab
Your weekend drink. No alcohol.
Window light
Each message is one flexible ad with two pictures, and Meta picks one for each person, so the result is about the message, not one picture. Nothing spends until you approve and start the ads.
Type
Concept test · 2 flexible ads
Split
Meta A/B test, one side per person
How long
Oct 17 – Oct 30 · 14 days
Judged by
Cost per add to cart
Can spot
A gap of 1.6× or more between the sides
Spend limitHard stop
$600 then $1,500 of $1,500 in Oct
What it will narrow
Illustrative
Swap vs taste53 → 44 pts
−20%No difference+20%+40%
Before your test: outside evidence
Expected after this test
How it works
Most teams start every new ad from scratch. Loopcraft keeps what each test taught, so the next ad starts from what works.
01
Connect your brand
Link your website, shop, ad accounts and reviews, and upload your past ads in bulk. Loopcraft reads why people buy, what stops them and which ads worked.
02
Find ideas
Save competitor ads and see how long each has run. Your own ads and their results sit beside them.
03
Upload or make ads
Upload the ads you already have, or let Studio make images and short videos from a brief. It changes only what a test compares, so the result shows what made the difference.
04
Run fair tests
You approve the plan, the ads and the spend limit. Each person sees one version, and nothing spends before you say yes.
05
Learn what works
Each result becomes a plain sentence about your customers, like “people try Sundays for the taste”, with how sure Loopcraft is.
06
Know what to make next
Loopcraft says which ads to make next and why, and which question is worth a test. It does not recommend paying to test picture details.
Find ad ideas.Ask what worked.
Search competitor ads and your own, ask why an ad won, and act on any result, all from one search bar.
Competitor ads: on Today, you press Command K and type “longest-running competitor ads”. Loopcraft shows five competitor ads from 41 brands you follow and the public ad library, sorted by how long each has run. You move to Solstice: it says “Real oranges, squeezed and fizzed. That’s the whole can.”, links to its orange soda with “Shop now”, and has been seen for 74 days since Aug 2, in 3 versions, on Facebook and Instagram. Loopcraft reads that it leads with the taste, so it counts toward “Lead with the taste”, weakly: running long hints that it works for them; it is not their results. You can make 2 drafts in that direction for 8 credits, plan a test from its idea, save it or find similar ads. The picture is an example for the fictional Sundays brand.
SSundaysExampleSearch ads, briefs or askK$900 of $1,500 in Oct1,240 creditsMO
Result to confirm1 / 6
The taste message got 25% more clicks. Confirm the reading?
Test 1 met its rule at its $900 limit on Oct 12. The taste won with one picture, so the belief is promising, not settled: Test 2 checks it again on two new pictures. Purchases went the same way, too few to judge alone.
Sundays, slow and sunny.
AMood message
Clicks
720
Views
45,000
Click rate
1.60%
Spent
$450
Purchases, illustrative
13
Real yuzu, lightly sweet.More clicks
BTaste message
Clicks
900
Views
45,000
Click rate
2.00%
Spent
$450
Purchases, illustrative
19
What Loopcraft believes now
Illustrative
Leading with the taste likely brings +11% to +31% more clicks than the mood.
+11%+20%+31%
−20%No difference+20%+40%
Before your test: outside evidence
Your test alone
Now: everything together
Range narrowed from 36 to 20 points
Why the range is narrower than your test alone
Your test counts for 74%. Three outside sources agreed, more cautiously:
Example benchmark · Flavour-led drink ads15%
Your account history · 10 past Sundays ads6%
Competitor ads · 12 long-running drink ads5%
If you confirm
“Lead with the taste” becomes promising for clicks
Customers marks the taste reason “Promising after Test 1”
Tests in Learn re-rank on the new range
longest-running competitor adsin SundaysK
5 ads · 41 brands you follow and the public ad librarySorted by days running · a hint, not results
SolsticeActive today
Fictional competitor · Facebook, Instagram
Real oranges, squeezed and fizzed. That’s the whole can.
Orange soda, 12 cansTaste the whole orangeShop now
Seen74 days since Aug 2 · 3 versions
JulAugSepToday
Leads with
The taste: it names the fruit first
Shows
Fruit splash · Close-up · Bold colour
Spend, results
Not public
Counts toward “Lead with the taste”, weaklyRebuilt twice and still active: a hint, not their results.
Do with Solstice
Make 2 drafts in this direction8 creditsD
Plan a test from its ideaFreeP
Save to libraryFreeS
↑↓moveopendo the highlighted action1,240 creditsEscclose
From idea to result
Make, test and learn in one place, with your approval before anything spends.
Brand kit
Sundays · read from your website and files
Sparkling yuzu soda, lightly sweet, made for slow weekend afternoons.
Adults 21 to 45 in the US who buy craft soda instead of beer or cocktails on weekends.
Colours
Fonts
Sundays, slow and sunny.
Fraunces SemiBold · Instrument Sans
Voice
Short headlines: five words or fewer.Added by Maya, Sep 28
Say “yuzu”, never “citrus blend”.Brand guide PDF, page 3
No health or wellness claims.Added by Maya, Sep 28
Saved · every brief and draft starts from it
Brand
Save your brand once.
Loopcraft reads your website and files and keeps what you sell, who buys it, your colours, fonts and voice. Every brief and draft starts from it, and nothing changes without your yes.
Customers
From 214 reviews · shares are illustrative
Website14 pages read
ShopOrders and first orders, since Sep 1
Reviews214 reviews
Meta ad account10 past ads and Test 1
Why people buy
The taste: real yuzu, not too sweetPromising after Test 146%
A weekend drink instead of alcoholNo ad says it yet · Test 231%
Less sugar than regular sodaNo ad says it yet18%
Customers
Start from why people buy.
Loopcraft reads your reviews, orders, website and past ads. It finds the reasons people buy and the doubts that stop them, and ties every test to one of them.
StudioTaste or mood?Brief · one change: the messagethe brief behind Test 1
Sundays, slow and sunny.
AMood messageTaste message
Sundays, slow and sunny.
Real yuzu, lightly sweet.
Headline
BMood messageTaste message
Sundays, slow and sunny.
Real yuzu, lightly sweet.
Only the message differs
Studio
Make ads that test one thing.
Studio makes images and short videos from a brief. To test a message, it keeps the picture and changes only the words.
Your ads
Uploading 12 ads12 ads read · tagged · results joined
Drop ads hereImages, videos or an export from your ad account
Real yuzu, lightly sweet.
Taste message
Taste4:52.00%
Window light
Taste4:51.90%
Fridge grab
Taste9:161.80%
Sundays, slow and sunny.
Mood message
Mood4:51.60%
3 selected· all taste messageTest themMake versions6 versions in Studio
Library
Upload all your ads at once.
Drop in past ads, videos or an export from your ad account. Loopcraft tags each one by message and format, joins its results, and lets you test or remake many at once.
Do more people add to cart when we sell the weekend swap instead of the taste?
Plan· waiting for your approval of $600
FromTest 1: taste beat mood, +11% to +31%67 of 214 reviews: a weekend drink instead of alcohol4 of 9 long-running competitor ads sell a swap
Test plans
Every test says what it is for.
Each plan names the question, where it came from and what each answer changes. Nothing spends until you approve, and each ad stops at its limit.
Done· the taste message got 25% more clicks
What Loopcraft believes now
Illustrative
Leading with the taste likely brings +11% to +31% more clicks than the mood.
+11%+20%+31%
−20%No difference+20%+40%
Before your test: outside evidence
Your test alone
Now: everything together
Range narrowed from 36 to 20 points
“Lead with the taste”· Promising for clicks · one picture
NextThe weekend swap or the taste?
Results
Learn something you can use again.
Each result updates what Loopcraft knows about your customers, like “people try Sundays for the taste”, with a range for how sure it is.
Works with your tools.Open to your agents.
Connect your ad accounts, store, files and team. Your AI agents can use Loopcraft through MCP.
Connect your system and agents with GraphQL or MCP
Claude
ChatGPT
Gemini
Muse
Cursor
Grok
Manus
Your agent
Meta
Instagram
TikTok
Google Ads
YouTube
Pinterest
Snapchat
LinkedIn
Google Analytics
Google Drive
Dropbox
Canva
Figma
Slack
Gmail
WhatsApp
Your tools
See all 91 connections
Ad platforms
Meta (Facebook, Instagram), TikTok, Google Ads, YouTube, Pinterest, Snapchat, LinkedIn, Reddit, X, Amazon Ads, Apple Ads
Measurement and attribution
Meta Conversions API, Meta Pixel, Google Analytics 4, Shopify Analytics, Triple Whale, Northbeam, Polar Analytics, Rockerbox, Elevar, AppsFlyer
Commerce and catalogs
Shopify, WooCommerce, BigCommerce, Amazon, Etsy, Meta Commerce catalog, Google Merchant Center
It learns what makes your customers buy and turns it into ads. It reads your brand and results, finds ad ideas, makes the ads, runs fair tests with your approval and tells you what to make next.
How does it learn?
It keeps what it knows as plain sentences about your customers, such as “people try Sundays for the taste”, each with a range for how sure it is and the evidence behind it: your tests, past ads, reviews and outside benchmarks. It tests what carries over to your next ads, like why people buy and what to say. It does not recommend paying to test picture details; it varies them inside each test.
My ad spend is small. Can Loopcraft still help?
Yes. It starts from what your customers already tell you in reviews and orders, then adds your past ads, long-running competitor ads and industry benchmarks, and shows how much each counts. With a small budget it runs fewer tests with bigger differences, so each result is clear.
Does Loopcraft make the ads too?
Yes, in Studio. It makes images and short videos from a brief or any ad, and changes only what a test compares. Every new version waits for your review.
Where can I launch?
On the ad accounts you connect: Facebook and Instagram, TikTok, Google and YouTube, Pinterest, Snapchat and LinkedIn. You set the spend limit, and every test needs your approval.
Can Loopcraft launch or spend without my approval?
No. Every test needs your approval of the amount, the ads and the plan, and each ad stops at its spend limit. Raising a limit always asks.
What does starting free include?
No Loopcraft charge for your first full test, from connecting your brand to a launched test and its result. No card is needed to start. You pay the ad platform directly for your ad spend.
Can I connect my own AI tools?
Yes, through MCP. Create an access key for one brand, then add it to Claude, ChatGPT, Gemini, Cursor or any other app that supports MCP. Keys can never publish, spend or approve.
Competitor adRunning 74 days
Test 2 planWaits for approval
Test 1 resultThe taste won
Studio draftThe weekend swap
“People try it for the taste”Learned
Run your first test free.Runyourfirsttestfree.
Your first full test costs nothing, and no card is needed. You pay the ad platform for ad spend, and nothing launches without your approval.
