Command Your CRO Program From Claude or ChatGPT
With the Shoplift API, you can plan, prioritize, schedule, launch, analyze, and report on every experience you run — across one store or your whole portfolio — driven by simple commands.
Your CRO program runs where your team already works
Import your roadmap in one command
Import your roadmap in one command
Queue the quarter in one sitting
Queue the quarter in one sitting
Ask at the level you need
Ask at the level you need
It runs where your team already works
It runs where your team already works
Run your multi-brand portfolio from one command line
Run your multi-brand portfolio from one command line
What you can do with it
Bring it in
Whether your roadmap lives in Google Sheets, JIRA, Notion, or a PDF, import it through the API with all its metadata, organized into Campaigns. And as you use AI across your business, feed hypotheses straight into Shoplift, drop drafts and ideas into the backlog, and come back later to turn them into a formal plan. All from the AI of your choice. Every import lands as an idea, never on your storefront.
Run it
Plan, prioritize, and schedule every experience without leaving your AI interface, with granular control over Audiences, Campaigns, A/B tests, and Personalizations. Launch a test the moment you think of it, schedule a full Spring-sale site overhaul weeks out, or fine-tune an experience that's already live.
Read it
Whether you want a status check on one test or the aggregated results and learnings of your entire CRO program for an executive review, you can do it from inside your AI. insights are backed by Shoplift’s Bayesian Statistical Engine to stress-test new hypotheses, plan follow-up tests, or pull insights from the data. And because your tests now have a memory, you can finally answer "did we test this before, and what did we learn?"
Orchestrate
Most testing tools stop at one storefront. If you run more than one — a portfolio of brands, a book of agency clients — the API lets you run all of them from one CLI. Launch one test everywhere at once. Aggregate results across the portfolio. Report across brands from one place. A key opens one store and only that store, so orchestrating the whole portfolio is easy and safe.
Hand it off
Point Hermes, Openclaw, or any other agent at your CRO program and let it work the backlog: read last week's results, form the next hypothesis, launch the test, report back.
How it works
One key, one reference, and your LLM runs the program.
Generate your key
Point your AI at it
Say what you want
Ask how it is going
Developer-Grade API Construction
From the only Shopify Plus-Certified CRO Platform
Endpoints
Authentication
Freshness
Validation
Rate limit
Parity
The only Shopify Plus Certified CRO platform
No. Connect your key in Claude Cowork or a CLI like Claude Code or Codex and describe the test you want in a sentence. If your engineers would rather write the calls, it's plain HTTP.
Yes. Point your AI at the Google Sheet, Notion database, PDF, or platform export you already have, and it creates every row as an experiment with the hypothesis, goal metric, and impact and effort scores intact. Results from another testing platform stay where they ran.
Settings, then API, inside Shoplift. Each key opens one store, and each store has one active key.
Experiments across their full lifecycle, their metadata, custom audiences, tags, campaigns where your plan includes them, and live results. Never customer records, order details or PII.
Yes, if you ask it to. Launch is an endpoint, the same as it is a button in the app. What bounds it: every test the API creates starts as a draft and stays there until something calls launch, a key reaches one store and only that store, and pausing a running test puts every visitor back on your normal storefront immediately. Nothing launches on its own unless you set a start date.
Generate a new one in Settings. The store's previous key stops working once the new one is active.
Template tests, theme tests, URL redirect tests, and two script-based types.
Yes. Set a future start and end date in your own timezone, and set a confidence threshold that stops the test when it gets there. That combination makes queueing a full quarter practical.
Everything the app shows. Probability of winning, lift, revenue impact, the state the test sits in and whether the read has held steady, plus CVR, AOV, RPV, CTR and ATC, split by device and by new versus returning, with day-by-day history.
60 requests per minute per key. Past that you get a 429 with a retry-after header.
One is coming in 2026, for the unstructured side of a CRO program — the reasoning, the session context, the why. Test data is already structured, so today your AI works from our reference and calls the API directly.
No. The API covers experiments.
All of them. Assigning experiments to a campaign requires a plan that includes Campaigns.
Yes, and it's one of the strongest reasons to use the API. Each store has its own key; loop your keys from a CLI and you can deploy one roadmap across every store, launch the same test in many stores or regions at once, and pull results back aggregated. A key only ever touches its own store.
The JavaScript API is a test type: custom logic, triggers and targeting running inside the page. This runs the program from outside the app. Different jobs, and teams use both.
