Audience-specific hypotheses
Grounded in a deep understanding of your business, your customers and what drives their decisions.
New: a landing page for every ad →
For conversion rate optimization (CRO) and experimentation teams
Your AI agent for A/B testing and personalization.
pagent finds where visitors drop off, builds tests in your shop's design and runs them to a result. You just approve.
Demo on your own website, plus free test ideas.
Used by
Run more A/B tests with less work from your team.
Your AI Experiment Engineer turns ideas into live tests. Choose what your team reviews.
Hypotheses grounded in your customers, their behaviour and your business. Start with your URL; add context as you go.
From copy to complex layouts: pagent builds real-code variations in your brand’s style, ready to preview on your site.
You approve every test before it goes live, in chat or by email.
Measure conversions with real visitors and Bayesian statistics. Each result informs the next hypothesis.
All 4 steps done
Experiment quality
One prompt. Ready for real visitors.
Built for your brand
Variants built in your design system and checked on desktop and mobile.
Feedback handled
Review by email. pagent incorporates changes and brings the next version back for approval.
15% average conversion uplift within 30 days.
Each test shows what works. pagent uses these findings to plan the next test, so the gains and lessons build over time.
E-Commerce+16%
Order Conversion Uplift
Text changes made with pagent raised Pixum's wall art conversion rate by +16%.
“With pagent, we established an AI-supported system that supplements our experimentation with fast, self-learning variants.”
Björn Prickartz
Head of Digital Analytics and Optimization, Pixum
Hospitality+80%
Click-Through Rate Uplift
A/B tests run with pagent raised the click-through rate for luxury hotels by +80%.
“Our clients operate in a highly competitive market environment. Our main mission is to help them gain a competitive advantage.”
Hannes Kinigadner
Team Lead SEA & Paid Social, Brandnamic
Setup & results
Fits your testing stack.
One script, added once
Keep your existing tools. Run pagent’s tests without a new deploy each time.
Results you can act on
Compare every variant against a control. See the lift, uncertainty and recommended next step.
Learning that carries forward
Break down results by audience and use each finding to shape the next hypothesis.
Claude & ChatGPT
Use pagent in Claude and ChatGPT.
Add pagent to the chat your team already uses. Ask which test won, turn an idea into a new test and read the results without opening another tool.
Owners and admins connect pagent in a few clicks and choose what the chat app can read and change.
Questions teams ask before a pilot
Can we keep our existing A/B testing tools?
Yes. pagent adds experimentation capacity alongside the tools you already use. Its script runs pagent’s variants on your site. During setup, we agree which pages and audiences pagent will work on and how that fits your existing tests.
What do we need to install?
Add pagent’s script tag to your website once. Variants are then delivered through that script, without a separate deploy for each test. We walk through the setup with you before the first test.
Can our team review changes before launch?
Yes. Your team can preview changes, approve, request revisions or veto by email, without a login. pagent incorporates the feedback. You choose which changes need review, or enable autopilot for the full cycle.
Can we use pagent in Claude or ChatGPT?
Yes. pagent is listed in Claude’s connector directory and ChatGPT’s plugin directory. Organisation owners and admins add it, sign in with their pagent account and choose what the chat app can read and change. They can then ask about results, capture hypotheses and start tests from the chat.
How are results measured?
Each test compares variants with a control against a primary conversion goal. Bayesian analysis reports lift and uncertainty. Results can support rolling out a variant, keeping the original or gathering more data. Each result informs the next hypothesis.
How much traffic do we need?
It depends on your conversion rate, the effect you want to detect and how many variants you test. We use your traffic and goals to scope a useful first test. Lower traffic or smaller effects can require more time.
How does pricing work?
The platform license is based on monthly unique users and includes all platform features plus a monthly allowance for agentic tasks. Additional task packs are optional. We agree the pilot scope and price with you before it starts.
How does pagent handle privacy?
pagent works with first-party data and does not rely on third-party cookies. We discuss your tracking and consent setup before the pilot. You can find more detail in our privacy policy.
What happens after we request a first test?
We contact you to understand your site, goals and current tools. Together we agree a focused pilot, pricing, setup and review settings. pagent then prepares the first experiment.
Let’s plan your first test.
Send us your website. We’ll contact you to agree the first test and its price.
We'll arrange a demo on your website.