Put 650+ A/B Tests Inside Your AI Assistant NEW

Connect the GoodUI MCP server to Claude or another AI assistant. Plan roadmaps, design stronger tests, and decide what to ship or iterate, based on what won, what lost, and your own custom meta-analyses.

See Team & Company Plans

Included with Team and Company-Wide plans. Already a member? Request your key.

What You Can Ask

Planning

Build A Testing Roadmap From Scratch

Turn the strongest evidence for your pages and metric into an ordered plan.

We sell subscriptions and have never tested checkout. Give us a 3-month test roadmap, prioritized by what has won most reliably.
Planning

Reprioritize Your Existing Roadmap

Re-rank the ideas you already have by how often similar changes have won.

Here is our backlog of 15 test ideas. Which have the strongest evidence, and which should we drop?
Before you test

Get Your Top 10 Design Ideas For Any Screen

Share a screenshot or URL and get the highest-impact changes, ranked by evidence.

Here's our pricing page. What 10 changes are most likely to lift signups?
Before you test

Raise The Odds Of Your Planned A/B Test

Check your treatment design against what has won and lost, before you build it.

We're about to test a shorter checkout form with trust badges. Which treatments should we double down on, and which should we avoid?
Before you test

Plan A Test Big Enough To Detect A Realistic Effect

Get the sample size, duration and MDE your test needs, based on the effects similar tests actually achieved.

We get 30,000 checkout visitors a week at 3% conversion. Using the meta-analyzed effects of past trust badge tests, how long should our test run?
After you test

Turn Finished Tests Into Next Steps

Get iteration ideas for a finished test, or find flat and losing tests worth another, better-powered try.

Our sticky CTA test was flat on add-to-cart. What should we try next, and is it worth rerunning?
After you test

Combine Your Own A/B Test With GoodUI's Evidence To Know Whether To Ship Or Iterate NEW

Add your test's visitors and conversions to similar GoodUI tests. You gain statistical power from their combined sample, see whether your result holds up beyond your own site, and get a clear next step: ship, retest or iterate. Same method as our pattern pages and meta-analysis calculator.

Here's our checkout urgency test: 99,000 / 188,000 vs 99,800 / 188,000 orders. Combine it with similar GoodUI tests.
ExampleVisitorsEffectp-value
GoodUI evidence (3 tests)42,900+2.2%0.0001
Your test376,000+0.8%0.0092
Combined → Ship418,900+1.1%< 0.0001

Illustrative figures.

Available On Team And Company-Wide Plans

Compare Plans

How It Works

Get your key

Team and Company members request a personal key. We send it to you, usually within a business day.

Connect your AI tool

Add GoodUI as a connector in Claude, or in any AI tool that supports remote MCP servers. It takes about two minutes.

Ask in plain language

Your assistant searches patterns and tests, ranks results and runs meta-analyses for you.

Common Questions

  • What is MCP?

    MCP (Model Context Protocol) is an open standard that lets AI assistants use outside tools and data. With GoodUI's MCP server connected, your assistant can look up GoodUI patterns and A/B tests directly instead of guessing from what it learned in training.

  • Which AI tools does it work with?

    Claude (desktop and web) via a custom connector, and other AI tools that support remote MCP servers with an API key in a request header. We'll send setup steps with your key.

  • Can I run my own meta-analysis?

    Yes. Ask your assistant to combine any 2 or more GoodUI tests, for example all tests of a pattern on one metric, or just the ones that match your page and device. You can include your own results too: give your assistant the visitors and conversions for A and B, and they're combined with the GoodUI tests using the same inverse-variance method as our pattern pages. You get one combined effect and p-value, plus each test's effect and weight. Your own figures are only used for the calculation; we don't store them.

  • Does GoodUI see my conversations?

    No. We only receive the requests your assistant sends to GoodUI, such as a search term or a test number, never your prompts or the rest of the conversation. We log those requests to run the service and keep usage fair; see our Terms of Service.

  • Are there usage limits?

    MCP access is for your team's own research and experiment planning, under fair use. Normal use, even heavy use, is fine. Bulk downloading or copying the dataset isn't. Details are in our Terms of Service.

  • Can we build GoodUI data into our own product or AI model?

    Not with a membership. That needs a data license, which is available: talk to Jakub.

Get Team Or Company Access