A/B tests
Test every change. Keep what wins.
Your agent picks the test from your own numbers, ships the new version as a pull request, and tells you plainly which one won. Free.
Works with Claude Code and Codex.
How it works
Your numbers pick the test.
- 01
You type it
One command in Claude Code, or ask Codex what to test.
- 02
Your agent builds the test
It proposes three changes with the numbers behind each, sizes the one you pick from your traffic, and builds it in your code.
- 03
You merge, then keep the winner
Merge to start the test. When the answer is in, one more pull request makes the winner your page.
> /shipfound:test> What should we A/B test?What you get
From idea to winner, in your code.
Ideas with evidence
Each change comes with the numbers from your own pages behind it: visits, how many stay, how many sign up.
Sized before it starts
How many visitors each version needs, and how many weeks that takes at your traffic.
Every visitor counts
Visitors are split with nothing stored in their browser, so everyone who lands is in the test.
A straight answer
The chance the new version beats your page, and by how much. A winner only when the numbers say so.
The winner shipped
One pull request makes the winner your page's only version and takes the test code out.
Free, on your own analytics
Tests read Shipfound's free analytics. Not on your site yet? Your agent adds them first, as one pull request.
FAQ
Questions founders ask first.
Stop guessing. Test it.
Works with Claude Code and Codex.