MCP

Let your agent run A/B tests over MCP

Connect your agent to pagent and run A/B tests on your live site from the chat you already use. pagent builds each variant, checks it on desktop and mobile, and runs the test with Bayesian statistics. Your agent reads the results and starts the next test.

Start free trialAdd pagent to Claude

Free for 14 days. New to MCP? Read the setup guide.

Example conversationpagent connected

Which tests are running on shop.example.com, and which one is ready to ship?

pagent · list_testsRead 3 tests

Two tests are running. Checkout with fewer fields is ready to ship.

Checkout with fewer fields+12%
Chance to beat control97%
Completed orders+12%

Should I capture the next hypothesis for /checkout?

Endpoint
https://app.pagent.ai/api/mcp
Transport
Streamable HTTP
Auth
OAuth 2.0 or personal token
Registry
io.github.aihumin/pagent

Agent-native

An agent your agent can use

pagent is an agent itself. It defines hypotheses, builds and checks variants in real code on your design system, and runs the tests on your live site. Your own agent can also use pagent as a tool. Claude, ChatGPT, Codex, Cursor or any other MCP client signs in over OAuth to read results, capture hypotheses, create and start tests, and manage audiences and goals.

Works with

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Zed

  1. Your agent sets the goal.

  2. pagent does the work.

  3. Your team reviews each variant before it goes live.

Capabilities

What your agent can do

With read access

  • List running and completed tests and read their Bayesian results in plain language
  • Compare tests and see what each variant changes on the page
  • Read pages, audiences, conversion goals, personas, strategies and guidelines
  • List the websites and organisations the connection can see

With write access

  • Capture a hypothesis and turn it into a test
  • Start, pause, resume, queue, stop, evolve or roll back a test
  • Create pages and conversion goals, and manage audiences and their pages
  • Update the personas, strategies, keywords and guidelines pagent works from

When you sign in over OAuth, you pick the access level and the organisations the client can see, for each connection. You can revoke a connection at any time in your settings.

Setup

Connect your agent

Claude

Open pagent in Claude's connector directory, select Connect, sign in, and choose organisations and permissions.

Add pagent to Claude

ChatGPT

Turn on Developer mode under Settings, Security and login. Open Plugins, select the plus button and name the plugin pagent. Enter the endpoint as the server URL, keep OAuth as the authentication, and select Create. Sign in with pagent, choose organisations and permissions, and allow access. To use it, open Plugins, pagent, Try in chat, or type /pagent in the composer.

Terminal and editors

Claude Code

claude mcp add --transport http pagent https://app.pagent.ai/api/mcp

Codex

codex mcp add pagent --url https://app.pagent.ai/api/mcp
codex mcp login pagent

Cursor

{
  "mcpServers": {
    "pagent": { "url": "https://app.pagent.ai/api/mcp" }
  }
}

VS Code, Windsurf, Zed and personal tokens: the setup guide.

Prompts

What to ask once connected

Name the website in every request. The connection does not remember it between messages.

  • Which tests are running on shop.example.com, and which one is ready to ship?
  • Turn this hypothesis into a test on /checkout: fewer form fields above the fold raise completed orders.
  • Compare the two hero tests from August and explain the result to my team.

The difference

pagent runs the whole test loop

Some testing MCP servers only create a test from selectors your agent writes. With pagent, your agent names the goal. pagent builds the variant in real code with your components and tone, then runs the test with Bayesian statistics, sequential correction and a sample-ratio check. Each winner becomes the new baseline and shapes the next hypothesis.

How the AI A/B testing agent works. Compare pagent with the best AI A/B testing tools in 2026.

  1. Reads the page
  2. Writes the hypothesis
  3. Builds the variant in real code
  4. Checks it on desktop and mobile
  5. Runs it with Bayesian statistics
  6. Makes the winner the new baseline

Frequently asked questions

Is there an A/B testing MCP server I can use with Claude or ChatGPT?

Yes. pagent runs a hosted MCP server at https://app.pagent.ai/api/mcp over streamable HTTP with OAuth 2.0. It is listed in Claude's connector directory and the official MCP registry. It works with Claude, Claude Code, ChatGPT, Codex, Cursor, VS Code, Windsurf, Zed and any other MCP client.

Can Claude run A/B tests on my live website?

Yes, with the pagent connector. Add pagent from Claude's connector directory, sign in, and ask Claude to turn a hypothesis into a test. pagent builds the variant in real code, checks it on desktop and mobile, and runs the test on your site with Bayesian statistics. Claude then reads the results and captures the next hypothesis.

Does my agent have to write the variant code?

No. Your agent gives the goal and the hypothesis. pagent writes the variant in real code on your design system, checks it on desktop and mobile, and ships it through the script tag already on your site. MCP servers that only create a test from selectors leave that work to your agent.

What is the difference between read and write access?

Read access lets an agent list tests, read Bayesian results, compare variants, and look at pages, audiences, goals and personas. Write access adds the test lifecycle: capture hypotheses, create and start tests, pause, stop, evolve or roll back, and manage audiences and goals. You pick the level for each connection when you sign in over OAuth.

Is pagent agent-native?

Yes, in both directions. pagent is an agent itself: it defines hypotheses, builds and checks variants in real code, and runs the tests on your live site. Your own agent can also drive pagent as a tool over MCP.

Does anything ship without a review?

No. Your team reviews every variant in the dashboard, by email, or in the chat you already use. pagent does not add its own confirmation step for MCP writes, so keep write confirmation on in your client and ask it to show planned changes before it acts.

What do I need before I connect?

A pagent account with owner or administrator access to the organisation, MCP turned on for that organisation, and an onboarded website with the pagent script installed. Without the script, there are no tests or visitor data to read.

Run your next A/B test from your agent

Start a 14-day free trial and connect your agent on day one.

Start free trialAdd pagent to Claude

Prefer a walkthrough? Get a demo