> ## Documentation Index
> Fetch the complete documentation index at: https://www.pagent.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Claude connector

> Connect Pagent to Claude to review website tests, capture hypotheses, and manage your optimization workflow.

Connect Pagent to Claude so you can inspect website test performance, compare what changed, capture new hypotheses, and manage permitted test actions from a conversation.

<Card title="Connect Pagent to Claude" icon="plug" href="https://claude.ai/directory/connectors/pagent" horizontal>
  Open the Pagent connector in Claude and select **Connect**.
</Card>

## What Claude can do with Pagent

* List running and completed tests and explain their Bayesian results
* Compare hypotheses, variations, audiences, and measured outcomes
* Capture hypotheses and turn them into tests
* Work with pages, audiences, conversion goals, personas, and guidelines
* Start, pause, resume, stop, or roll back tests when you grant write access

The tools Claude sees depend on the permissions you approve. Read access and write access are kept separate.

## Before you connect

You need:

* A Pagent account with owner or administrator access to the organisation you want to connect
* MCP enabled for that organisation
* An onboarded website with the Pagent SDK installed if you expect visitor data or test results

## Connect Pagent

Select **Connect Pagent to Claude** above, then sign in to Pagent. Choose the organisations Claude may access and approve the permissions you need. Claude handles the connector configuration for you.

Once connected, Pagent is available anywhere directory connectors are supported in Claude.

## Authentication and permissions

Pagent uses OAuth when you connect from Claude. During approval, you choose:

* Which organisations the connection can access
* Whether the connection remains read-only
* Which optional write permission groups to enable

Pagent checks your current organisation membership, role, the connection's selected organisations, and the organisation's MCP policy on every request.

<Warning>
  A write runs when Claude calls the tool. Pagent does not add a second
  confirmation screen. Review Claude's proposed action before approving a write,
  especially when it can affect a live test.
</Warning>

A connection does not remember a website between requests. Name the intended website or domain when asking Claude to perform website-specific work.

## Try these prompts

### Review live tests without changing anything

```text theme={"dark"}
Across every website this connection can access, list the running tests. For each test, report the page, audience, views, result status, lift, probability to beat control, and Pagent recommendation. Treat inconclusive results as inconclusive. Do not change anything.
```

Claude reads the available test and result data and returns a report without invoking write tools.

### Compare two tests

```text theme={"dark"}
On [website], compare tests #12 and #18. Show the hypothesis, the concrete page changes in each variation, and the measured result. Identify the strongest learning, but do not create a follow-up yet.
```

Claude retrieves both tests and explains the differences and measured learning.

### Turn a hypothesis into a test

```text theme={"dark"}
On [website], capture this hypothesis for /pricing: [hypothesis]. Show me the exact hypothesis, target page, and next steps first. After I confirm, create the test and tell me when it is ready for review.
```

Claude shows the planned write first. After your confirmation, it can create the hypothesis and test if the connection has the required permissions.

### Check before starting a live test

```text theme={"dark"}
Check test #42 against its current lifecycle, review status, SDK status, and any conflicting test. Explain what starting it would change and wait for my confirmation. Start it only after I explicitly say start.
```

Claude validates the current state before proposing a lifecycle action.

## Manage or revoke access

Open [AI client access](https://app.pagent.ai/me/settings/mcp) to review connections, change organisation access, see when a connection was last used, or revoke it. Reconnect Claude if you need to change its permission groups.

## Troubleshooting

<AccordionGroup>
  <Accordion title="No websites appear">
    The connection can only see organisations selected during approval. Reconnect or update its organisation access in Pagent. The organisation must also have MCP enabled.
  </Accordion>

  <Accordion title="A tool is missing">
    Pagent only exposes tools allowed by the connection's permissions. Reconnect
    with the required permission if you intentionally need broader access.
  </Accordion>

  <Accordion title="Claude cannot find a website">
    Ask Claude to list the available websites, then refer to the intended website
    by name or domain. Connections do not keep a selected website between
    requests.
  </Accordion>

  <Accordion title="An action was denied">
    Pagent checks your current organisation role and policy on every request. Your
    role may have changed, MCP may have been disabled, or the test may not allow
    that action in its current state.
  </Accordion>

  <Accordion title="OAuth failed or expired">
    Disconnect Pagent in Claude, add it again, and complete the browser approval flow.
  </Accordion>
</AccordionGroup>

## Privacy and support

Data returned through the connector is processed by Claude under Anthropic's terms and privacy practices. Review [Pagent's privacy policy](https://www.pagent.ai/privacy-policy) for how Pagent handles MCP connections.

For help, email [support@pagent.ai](mailto:support@pagent.ai) with the Claude surface you used, the approximate error time, and the exact message. Never send your personal token.

<Card title="Full MCP reference" icon="book-open" href="/docs/guides/mcp">
  See every supported AI client, capability, safety detail, and troubleshooting
  step.
</Card>