Pocketsflow

Sell moreA/B Testing

Let your buyers pick the winner.

Show each visitor one version of a product: a price, a headline, a cover or a checkout layout. Pocketsflow counts who buys, tells you in plain words when one version is clearly ahead, and one click makes it your product.

You can use your current AI to start an A/B test.

Sign in, click Allow. No API key.
Price test on Notion Business OS, running for 14 days: B is winning with a 97% chance to earn more per visitor and a Declare B the winner button, 6,800 visitors, 225 sales and $12,115.00 revenue, and a table with A (Current, $49.00) at $1.67 per visitor and B (Higher price, $59.00) at $1.90Price test on Notion Business OS, running for 14 days: B is winning with a 97% chance to earn more per visitor and a Declare B the winner button, 6,800 visitors, 225 sales and $12,115.00 revenue, and a table with A (Current, $49.00) at $1.67 per visitor and B (Higher price, $59.00) at $1.90
One test, every numberVisitors, sales, conversion and revenue per visitor for each version.
things to test: price, headline, cover, checkout layout
4
versions per test, A plus B, C and D
Up to 4
chance to beat A before a version is called the winner
95%
MCP tools for A/B tests
7

// inside pocketsflow

From idea to winner in the dashboard.

01Pick what to testA price, a headline, a cover image or a checkout layout, on one product.
New A/B test form with Price selected among Price, Headline, Cover image and Checkout layout, the product Notion Business OS at $59.00, version A $59.00 and version B $69 (+17% vs A), a 50/50 split in the Your test summary, and How it works notesNew A/B test form with Price selected among Price, Headline, Cover image and Checkout layout, the product Notion Business OS at $59.00, version A $59.00 and version B $69 (+17% vs A), a 50/50 split in the Your test summary, and How it works notes
02Watch it settleEach version's conversion rate day by day, with every number explained in plain words.
The versions table of a price test (A $1.67 and B $1.90 per visitor, B at a 97% chance to earn more and Winning), the Conversion rate over time chart for A and B from Oct 3 to Oct 16, and How to read this: Chance to earn more, vs A and Revenue per visitorThe versions table of a price test (A $1.67 and B $1.90 per visitor, B at a 97% chance to earn more and Winning), the Conversion rate over time chart for A and B from Oct 3 to Oct 16, and How to read this: Chance to earn more, vs A and Revenue per visitor
03Keep the winner in one clickThe winning version becomes your product for everyone, and the test ends.
Declare B the winner? dialog: $59.00 becomes your product's price for every visitor, and the test ends. So far: 3.22% conversion and $1.90 per visitor, with Cancel and Apply B buttonsDeclare B the winner? dialog: $59.00 becomes your product's price for every visitor, and the test ends. So far: 3.22% conversion and $1.90 per visitor, with Cancel and Apply B buttons
04Every test in one listRunning, won and stopped tests with their visitors, sales and conversion.
A/B tests page: Let your buyers pick the winner, 13,810 visitors tested, 524 sales counted and 3 tests, a preview of versions A ($49.00) and B ($59.00) where B earns 14% more per visitor, and the test list: a running price test on Notion Business OS (6,800 visitors, 225 sales, 3.31%), a headline test with a declared winner and a stopped checkout layout testA/B tests page: Let your buyers pick the winner, 13,810 visitors tested, 524 sales counted and 3 tests, a preview of versions A ($49.00) and B ($59.00) where B earns 14% more per visitor, and the test list: a running price test on Notion Business OS (6,800 visitors, 225 sales, 3.31%), a headline test with a declared winner and a stopped checkout layout test

// more screens

// use cases

What sellers test first.

  • Find your price

    Is the higher price worth it? Price tests are judged on revenue per visitor, not on sales alone.

    Notion Business OS$49 vs $59

  • Three price points

    Put up to four prices against each other and keep the one that earns the most.

    Icon Pack$19 · $29 · $39

  • A sharper headline

    Try a new product name, with its own description if you like.

    Street Photography in 30 Days$79

  • A cover that stops the scroll

    Test a new cover image on the product page and at checkout.

    Lightroom Presets Vol. 2$39

  • The right checkout

    Split against Focus, or any two of the checkout layouts.

    The Creator Funnel Course$149

  • Before a big launch

    Run the test on today's traffic, then launch with the version that won.

    Client Onboarding Kit$29

// how it works

Live in six steps.

  1. Pick what to test

    A price, a headline, a cover image or a checkout layout.

  2. Choose the product

    Any paid product. One test runs per product at a time.

  3. Add your versions

    A is your product as it is today. Add B, and C and D if you like; each must differ.

  4. Split the traffic

    Even by default, or a custom split with 5 to 95% per version.

  5. Read the results

    Visitors, sales, conversion, revenue per visitor and the chance each version beats A.

  6. Keep the winner

    One click applies the winning version to your product for every visitor.

// ai

Let your agents run and grow your business.

Connect Claude, ChatGPT, Codex, Grok, Cursor or any MCP client once. Then ask in plain words, and every change lands in your dashboard.

Claude · Desktop app

Things to ask

  • Test $59 against my current $49 pricelist_products · create_ab_test
  • Which version of my price test is ahead?get_ab_test_results
  • Declare B the winnerdeclare_ab_test_winner
  • Which tests are running right now?list_ab_tests
  • Show me the versions in my cover testget_ab_test
  • Stop the headline test on my presetslist_ab_tests · stop_ab_test
  • Delete the old checkout layout testdelete_ab_test

// included

Everything included.

  • Price, headline, cover image and checkout layout tests
  • Up to four versions: A plus B, C and D
  • Even or custom traffic split, 5 to 95% per version
  • One version per visitor, on the page and at checkout
  • Price set by the server at checkout
  • Refunds, bots and link previews left out
  • Conversion rate and revenue per visitor
  • Chance to beat A, or to earn more on price tests
  • Daily conversion chart for every version
  • Winner applied to the product in one click
  • Editing what you test stops the test
  • ab_test.completed webhook, API and MCP tools

// why pocketsflowWhy sellers pick Pocketsflow.

  • Judged on money

    A lower price can sell more often and still earn less, so price tests compare revenue per visitor.

  • Fair to buyers

    Each visitor sees one version and pays the price they saw. Nothing tells them a test is running.

  • Verdicts in plain words

    Too early to call until there is enough data, then winning, losing or no clear difference.

  • Results you can trust

    Refunds and bots are left out, and a test stops on its own if you change what it tests.

// connect

It's easy. Even your AI can do it.

Add Pocketsflow to the AI you already use, sign in and click Allow access. No API key, no code.

  1. Paste into your AI app
    api.pocketsflow.com/mcp
    It's a connector address, not a web page.
    01

    Add Pocketsflow

    Paste the server URL as a connector in your AI app.

  2. 02

    Sign in and allow

    Your AI opens Pocketsflow. Click Allow access.

  3. 03

    Ask for anything

    Products, orders, refunds, emails, in plain words.

# Variant A is the product as it is now; send the challengers.
curl -X POST https://api.pocketsflow.com/ab-tests \
  -H "Authorization: Bearer pk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "productId": "665f1c2a9b3e4a0012ab34cd",
    "type": "price",
    "variants": [{ "label": "Higher price", "price": 59 }]
  }'

# Live results, then keep the winner:
# GET  /ab-tests/{id}
# POST /ab-tests/{id}/winner  { "variantId": "..." }

Run tests from your own code.

Start a test, read live results and keep the winner over the API or MCP. The signed ab_test.completed webhook tells your server when a test ends, with every version's results.

Every resource is an HTTPS endpoint under one base URL with Bearer API keys. Live and test keys are scoped to their own data; the OpenAPI spec is served at /docs.json.

npm install pocketsflow

// faq

Common questions.

What can I test?

The price, the headline (the product name, and optionally its description), the cover image or the checkout layout of one product. Version A is your product as it is today, and you can add up to three more versions: B, C and D.

Can a visitor see two prices?

No. Each browser gets a random visitor id in a first-party cookie on your store, and that visitor keeps one version on the product page, at checkout and when they come back. At checkout the server charges the version from its own record, never from the page, so the buyer pays the price they saw.

What counts as a conversion?

A paid sale, credited to the version the buyer saw. Refunded sales are left out, and bots, crawlers and link previews never see a test and are never counted.

How is the winner picked?

Each version shows its chance to beat A, from the sales so far; price tests show the chance to earn more per visitor instead. At 95% the version is marked as winning. Until every version has 100 visitors and the versions have 10 sales between them, the test reads Too early to call. Nothing changes until you declare a winner: then that version becomes your product for everyone and the test ends.

What if I edit the product during a test?

Editing what you test (the price in a price test, the name in a headline test, the cover in a cover test, the layout in a layout test) stops the test on its own, so the results never mix two versions. Other edits are fine.

Do discount codes and flash sales still work?

Yes. In a price test the tested price is the list price: early-bird steps, flash sales, fair pricing, discount codes and personal deadlines apply to it the same way they apply to your own price.

Which plan includes A/B tests?

Starting a test needs the Pro plan. If your plan no longer includes A/B tests, every visitor sees version A until you upgrade, and the test resumes where it was. Reading results, stopping a test, declaring a winner and deleting a test always work.

Can my AI do this for me?

Yes. Add the Pocketsflow MCP server to Claude, ChatGPT, Codex, Grok, Cursor or any MCP client, sign in and click Allow access. No API key. Your AI can then create products, set prices and discounts, open checkouts, read orders and send emails, and every change shows up in your dashboard.

Open your store today.

Sell your first product in minutes, keep more of every sale, and let the AI you already use run the busywork.

Or start from the AI you already use