# pagent > pagent is your AI agent for A/B testing and personalization. One prompt to turn an idea into an on-brand variation ready for testing. pagent runs your experiments end to end. Review and approve with your whole team. pagent is built by aihumin GmbH in Bonn, Germany. GDPR compliant. No code required. ## Proof 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. - Brandnamic (Hospitality): Intelligent Messaging for Luxury Hotels: AI-Powered A/B Testing at Brandnamic. Click-Through Rate: 6.35% → 11.65%; Uplift: +80%; Test Duration: 2 weeks. https://www.pagent.ai/case-studies/brandnamic - NEOH (E-Commerce): How NEOH introduced A/B testing for the first time using AI and achieved a +25.74% revenue uplift in 7 days. Revenue Uplift: +25.74%; Performance Increase: +16.81%; Test Duration: 7 days. https://www.pagent.ai/case-studies/neoh - Pixum (E-Commerce): How Pixum achieved a +16% conversion uplift through AI-powered optimization and personalization within 3 weeks. Conversion Rate Uplift of: +16%; From Creation to Live Test in: 3 Days; Positive Test Results: 3x Faster. https://www.pagent.ai/case-studies/pixum - Zahnklinik Wien Döbling (Healthcare): How Zahnklinik Wien Döbling strengthens trust and increases conversion rate by +15% through AI-powered website personalization. Conversion Rate Uplift: +15%; Test Duration: 3 weeks; Automation Level: >90%. https://www.pagent.ai/case-studies/zwd ## How it works 1. Find test ideas: pagent reviews your site and user behavior, and returns test ideas specific to each page, ranked by expected impact and effort. Just add your URL. No data source needed. 2. Build each version: pagent builds any kind of test, from a simple copy change to complex variations, within your brand and design system. Deployed with one script tag: real code on your live site, no flicker. 3. Review by email: Your team approves, requests changes, or vetoes: by email, no login. Or let pagent run the full test on autopilot. 4. Test and learn: Real users, Bayesian statistics, weak versions stopped early. Winners become your new baseline and shape the next test. ## Product - Homepage: https://www.pagent.ai/ - AI Experiment Engineer and testing-stack fit: https://www.pagent.ai/ai-ab-testing-agent - Pricing: https://www.pagent.ai/pricing - Get started / demo: https://www.pagent.ai/get-started - Case studies: https://www.pagent.ai/case-studies - MCP for Claude and other agents: https://www.pagent.ai/docs/guides/mcp - Markdown index: https://www.pagent.ai/llms-full.txt - Docs index for AI crawlers: https://www.pagent.ai/docs/llms.txt - Full docs index: https://www.pagent.ai/docs/llms-full.txt - Tool guides (Kameleoon, Optimizely, VWO, AB Tasty): https://www.pagent.ai/guides - Best AI A/B testing tools 2026, by use case: https://www.pagent.ai/guides/best-ai-ab-testing-tools-2026 ## MCP - Endpoint: https://app.pagent.ai/api/mcp - Transport: Streamable HTTP - Auth: OAuth 2.0 or personal token (/me/settings/mcp) - Registry name (once published): io.github.aihumin/pagent ## Who it is for From e-commerce to hospitality: teams run their website experimentation on pagent, with measurable results. ## Optional - Blog: https://www.pagent.ai/blog - Changelog: https://www.pagent.ai/changelog - Solutions: https://www.pagent.ai/solutions - Tools: https://www.pagent.ai/tools - About: https://www.pagent.ai/about-us - Contact: https://www.pagent.ai/contact - Traditional A/B testing comparison: https://www.pagent.ai/vs-traditional-ab-testing