> For the complete documentation index, see [llms.txt](https://docs.elevateab.com/elevate-helpcenter/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.elevateab.com/elevate-helpcenter/mcp/mcp-overview.md).

# MCP Overview

Connect Elevate A/B Testing to Claude, ChatGPT, Cursor, VS Code, Gemini CLI, or any MCP client and ask about your Shopify experiments in plain language.

MCP lets you connect Elevate A/B Testing to an AI assistant, such as Claude or ChatGPT, so you can ask about your Shopify experiments in plain language. Instead of opening a report and reading it yourself, you ask something like "which experiments are running right now, and is any of them close to a winner," and the assistant answers from your store's live data.

Every connection starts read-only: the assistant can see your tests and results, and it can't change anything in your store. You can also let a connection create, launch, pause, resume and end experiments. See [Let Your AI Assistant Make Changes](https://docs.elevateab.com/elevate-helpcenter/mcp/let-your-ai-assistant-make-changes.md).

***

### What is MCP?

MCP stands for Model Context Protocol. It is an open standard that lets an AI assistant connect to an outside service and read data from it directly.

Without MCP, an AI assistant only knows what you paste into the chat. You would have to export a report, copy the numbers in, and ask your question. With MCP, the assistant queries Elevate A/B Testing itself, so it always sees your current data and you never copy anything.

### Which AI assistants can connect to Elevate A/B Testing?

Elevate A/B Testing has a setup guide for seven named clients, plus a general setup for any other MCP client.

| Client                                                      | How it signs in                            | Setup guide                                                                                                           |
| ----------------------------------------------------------- | ------------------------------------------ | --------------------------------------------------------------------------------------------------------------------- |
| Claude Desktop                                              | Log in with Elevate                        | [Connect Claude Desktop, Web, or Mobile](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-claude-desktop.md) |
| Claude Web and Mobile (claude.ai and the Claude mobile app) | Log in with Elevate                        | [Connect Claude Desktop, Web, or Mobile](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-claude-desktop.md) |
| ChatGPT                                                     | Log in with Elevate                        | [Connect ChatGPT](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-chatgpt.md)                               |
| Claude Code                                                 | Log in with Elevate, or an API key in CI   | [Connect Claude Code or Cursor](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-claude-code-or-cursor.md)   |
| Cursor                                                      | API key                                    | [Connect Claude Code or Cursor](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-claude-code-or-cursor.md)   |
| Gemini CLI                                                  | API key                                    | [Connect Gemini CLI](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-gemini-cli.md)                         |
| VS Code (GitHub Copilot in agent mode)                      | API key                                    | [Connect VS Code](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-vs-code.md)                               |
| Any other MCP client                                        | Log in with Elevate (OAuth 2.1) or API key | [Connect Other MCP Clients](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-other-mcp-clients.md)           |

### Where to find MCP in Elevate A/B Testing

Select **MCP** in the sidebar. It sits in the **Setup** group, between **Integrations** and **Custom events**. Old links to **Settings**, then **MCP**, take you there too.

The MCP page shows your connection status at the top, then a card for each client in two groups: **Chat apps** (Claude Desktop, Claude Web & Mobile, ChatGPT) and **Coding tools** (Claude Code, Gemini CLI, Cursor, VS Code, Other MCP clients). Select a card to see the exact steps for that client, with the server URL and any configuration ready to copy. The steps name your store, for example `your-store.myshopify.com`.

Every client uses the same server URL:

```plaintext
https://elevateab.app/api/mcp
```

The transport is Streamable HTTP. Elevate A/B Testing does not serve SSE.

### Two ways a client signs in to Elevate A/B Testing

**Log in with Elevate.** Claude Desktop, Claude Web and Mobile, ChatGPT, and Claude Code use this. You add the server URL to the client, log in to Elevate A/B Testing if asked, and approve read-only access for your store. There is no key to copy or store. The first time you ask the assistant to make a change, it asks you to approve again, with an **Allow changes** option.

**API key.** Cursor, Gemini CLI, VS Code, and some other clients send a key in an `Authorization` header. On that client's setup screen, select **Create key** and Elevate A/B Testing adds the key to the setup for you. Copy the setup straight away: for your security, the full key is only shown once. Each key belongs to one connection, so you can disconnect one tool without affecting the others. A key is read-only unless you tick **Can make changes** when you create it.

MCP keys are separate from the External API key, which has its own **API** tab in **Settings**. See [Authentication](https://docs.elevateab.com/elevate-helpcenter/developer-reference/external-api/authentication.md).

### How do I know the connection is working?

Each client's setup screen has a status box. It reads **Waiting for** the client, for example **Waiting for Claude**, until the first request arrives, and it changes on its own once you finish the steps. It then reads **Connected**, with the time of the last request.

On the main MCP page, the top card reads **Not connected yet** until any client connects, then **Connected**. Each connected client's card shows a **Connected** badge.

To test a new connection, start a new chat or session and ask: What tests am I running?

### Where to see and disconnect connections

Once a client is connected, the MCP page lists every connection under **Connections**: each AI assistant that logged in and each API key, with when it was added and when it was last used. Disconnecting one doesn't affect the others. See [Manage MCP Connections](https://docs.elevateab.com/elevate-helpcenter/mcp/manage-mcp-connections.md).

### What the assistant can access

The Elevate A/B Testing MCP server gives every connection 19 tools that only read data. None of them creates, changes, or deletes anything in Elevate A/B Testing or in Shopify.

| Tool                           | What it returns                                                                                                                                               |
| ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `list_tests`                   | Your store's experiments, with optional filters                                                                                                               |
| `get_test`                     | Full details of one experiment, including every variation                                                                                                     |
| `get_test_results`             | Visitors, conversions, orders, revenue, profit, and lift for each variation, optionally narrowed by date range and the same filters as the results page       |
| `get_statistical_significance` | The significance status for each metric, recalculated on every request, with a verdict only once the experiment has enough data, the same as the results page |
| `get_test_timeseries`          | One metric over time for each variation, the data behind the performance chart                                                                                |
| `get_test_segmentation`        | Metrics by visitor segment: device, new or returning, organic or paid, and traffic source                                                                     |
| `get_test_products_breakdown`  | Results product by product, with revenue, orders, units, and conversion rate for each variation                                                               |
| `get_test_engagement`          | Session duration, where visitors go next, bounce rate, or click rates on elements                                                                             |
| `get_test_orders`              | Orders attributed to the experiment, with the same detail as the raw data export, including line items                                                        |
| `get_test_events`              | Individual events recorded for the experiment, with device, browser, page, location, and traffic source                                                       |
| `get_test_event_count`         | How many events match a given set of filters                                                                                                                  |
| `get_test_filter_options`      | The browsers, devices, countries, sources, pages, and UTM values that actually appear in the experiment's traffic                                             |
| `get_test_history`             | The experiment's activity timeline, such as started, paused, and settings changed                                                                             |
| `get_store_overview`           | Your store's plan, currency, monthly revenue, number of running experiments, and total experiments run                                                        |
| `get_store_conversion_funnel`  | Your store-wide funnel: page view, add to cart, checkout, purchase                                                                                            |
| `get_store_analytics`          | A store-wide report: web summary, period comparison, top pages, traffic sources, product performance, or cart analysis                                        |
| `search_products`              | Your store's products, found by name                                                                                                                          |
| `get_product_media`            | A product's images, in order                                                                                                                                  |
| `get_product_variants`         | A product's variants                                                                                                                                          |

A connection you allow to make changes also gets 8 tools that create, launch, pause, resume and end experiments. See [Let Your AI Assistant Make Changes](https://docs.elevateab.com/elevate-helpcenter/mcp/let-your-ai-assistant-make-changes.md).

You do not call these yourself. Ask your question in plain language and the assistant picks the tools it needs.

Connections and MCP API keys are created and disconnected only in Elevate A/B Testing, never through the assistant.

### Is connecting an AI assistant to Elevate A/B Testing safe?

**Connections are read-only unless you allow changes.** A read-only assistant reads your Elevate A/B Testing data and nothing more: it cannot end an experiment, change traffic allocation, or edit a setting even if you ask it to. No connection, read-only or not, has access to your Shopify admin, your customer records, or your storefront code.

**You approve each connection.** Clients that log in with Elevate show an approval screen for read-only access to your store before they can read anything, and ask again, with **Allow changes** off, before they can make a change. Every connection then appears in the **Connections** list, where you can disconnect it at any time.

**Treat an MCP API key like a password.** Anyone holding it can read your experiment data, and change experiments if it was created with **Can make changes**. Elevate A/B Testing stores only a fingerprint of each key, never the key itself, which is why the full key is only shown once. Keep config files that contain a key out of shared repositories. If a key is exposed, disconnect it and create a new one.

### Terminology note

The app labels this feature and its data using the word "test" in places where this documentation uses "experiment." They refer to the same thing. When you ask your assistant about your experiments, it will find them.

### FAQs

#### Does connecting an AI assistant affect my running experiments?

Not unless you allow it. A read-only connection does not change traffic allocation, pause or start anything, or alter what visitors see on your store. A connection you allowed to make changes can launch, pause, resume and end experiments when you ask it to.

#### Which Elevate A/B Testing plans include MCP?

All of them. MCP is available on every Elevate A/B Testing plan, and every plan reaches it the same way, from **MCP** in the sidebar.

#### Can I use Elevate A/B Testing MCP on my phone?

Yes, in the Claude mobile app. Connectors belong to your Claude account, so once you add Elevate in Claude Desktop or on claude.ai, it works in the Claude mobile app too, with nothing to set up on your phone. See [Connect Claude Desktop, Web, or Mobile](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-claude-desktop.md).

#### Can several people on my team connect?

Yes. Each person sets up their own client, and each connection appears separately in the **Connections** list. Disconnecting one person's connection doesn't affect anyone else's.

#### I manage several stores. How do I connect each one?

Each connection reads one store. When a client logs in with Elevate, the approval screen grants read-only access to the store your browser is logged into, and there is no store picker. To connect another store, open Elevate A/B Testing from that store's Shopify admin in the same browser, then start the connection again from your AI tool.

Reconnecting the same connector switches it to the new store. To keep several stores connected at once, add a separate connector for each store, with a name such as Elevate Store B. Check that your AI tool allows more than one connector with the same URL.

API keys also belong to one store. For key-based clients, create a key in each store and add one entry per store, each with that store's key.

#### I already use the Elevate desktop extension in Claude Desktop. What should I do?

Nothing, unless you want to switch. The extension keeps working. To switch to the connector, remove Elevate A/B Testing under **Settings → Extensions** in Claude Desktop, then follow [Connect Claude Desktop, Web, or Mobile](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-claude-desktop.md).

***

### Next steps

* [Connect Claude Desktop, Web, or Mobile](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-claude-desktop.md)
* [Connect ChatGPT](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-chatgpt.md)
* [Connect Claude Code or Cursor](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-claude-code-or-cursor.md)
* [Connect Gemini CLI](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-gemini-cli.md)
* [Connect VS Code](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-vs-code.md)
* [Connect Other MCP Clients](https://docs.elevateab.com/elevate-helpcenter/mcp/connect-other-mcp-clients.md)
* [Manage MCP Connections](https://docs.elevateab.com/elevate-helpcenter/mcp/manage-mcp-connections.md)
* [Let Your AI Assistant Make Changes](https://docs.elevateab.com/elevate-helpcenter/mcp/let-your-ai-assistant-make-changes.md)
* [Reports Overview](https://docs.elevateab.com/elevate-helpcenter/analyze/reports/reports-overview.md): understand the data your assistant is reading
