Setting up Adverity MCP

On this page

This guide explains how to connect Adverity to your AI tool using MCP.

For an overview of available tools and use cases, see Using Adverity MCP.

Introduction

How does my AI tool sign in to Adverity?

When you connect an AI tool, it opens the Adverity MCP sign-in page in your browser. On this page, you choose one of two ways to sign in:

Sign in with Adverity: You sign in to your Adverity instance, select a workspace and the permissions the AI tool gets, and Adverity creates a Connect API key for this connection. This is the default.
Use your own API key: You enter a Connect API key that you created yourself, for example a key that lasts longer than 30 days.

What can the AI tool access?

With Sign in with Adverity, the AI tool gets the permissions you select at sign-in, in the workspace you select and its child workspaces. It cannot get more permissions than you have there when you sign in. With your own key, it gets the permissions of that key.

Prerequisites

Before connecting, ensure you have all of the following:

Adverity account

An active Adverity account with access to the workspace you want to connect.

Adverity instance URL

Your dedicated environment address, such as https://yourcompany.adverity.com.

Compatible MCP client

An AI tool account that supports MCP connections:

Claude: Requires a Claude Pro or higher plan. On Team and Enterprise plans, a workspace administrator controls whether members can add custom connectors. If Add custom connector is not available to you, use Connecting with Claude Code instead.
ChatGPT: Requires a Business or Enterprise plan. Note that initial setup may require workspace administrator privileges to enable developer mode for custom connectors.

Connect API key (only to connect with your own key)

A valid Connect API key generated from your Adverity instance. The key permissions determine what actions AI can perform. For more information, see Creating a Connect API key in Adverity.

Signing in to Adverity MCP

Every AI tool uses the same Adverity MCP sign-in page. The sections below for each AI tool explain how to open it.

At the top of the sign-in page, check the name and address of the AI tool. If the page shows UNVERIFIED EXTERNAL APPLICATION, Adverity has not verified this address. Continue only if you started this connection yourself and recognize the address.

Signing in with your Adverity account

To sign in with your Adverity account, follow these steps:

On the Adverity MCP sign-in page, in Your Adverity instance URL, enter the URL of your Adverity instance.
Select the I accept Adverity’s Terms of Use checkbox.
Click Sign in with Adverity. Your browser opens your Adverity instance.
If you are not signed in to Adverity in this browser, sign in.
On the Connect to MCP page, check the name of the AI tool and, if the page shows it, the address the connection is handed to. Continue only if you started this connection yourself.
In Workspace, select the workspace the AI tool can access. The AI tool can also access the child workspaces of this workspace.
In Permissions, select what the AI tool can do:

Read and write
The AI tool can view and change data, within your own permissions.

Read-only
The AI tool can view data, but cannot change anything. This is the default.

Custom
Select the permissions one by one.

If Permissions shows a list of areas instead of these options, each area starts at the most you can grant. Lower any area the AI tool should not change. If the page does not show Permissions, the AI tool gets all the permissions you have in this workspace.
Click Allow access.

As a result, Adverity creates a Connect API key for this connection, and your browser returns to your AI tool.

Connecting with your own API key

Use your own Connect API key if you want to choose the key permissions and expiry yourself, if you already have a key for this purpose, or if your Adverity instance does not support signing in with Adverity.

To connect with your own API key, follow these steps:

On the Adverity MCP sign-in page, in Your Adverity instance URL, enter the URL of your Adverity instance.
Select the I accept Adverity’s Terms of Use checkbox.
Click Use your own API key. If the page shows the API key form already, skip this step.
In Adverity API Key, enter your Connect API key.
Click Connect with an API key.

As a result, your AI tool is connected to Adverity with the permissions of your key.

Connecting to Claude

Claude offers two ways to connect to Adverity: the Connectors settings in Claude.ai, and the Claude Code command line. Use Claude Code if custom connectors are not available in your workspace, or if you want to connect more than one Adverity instance at the same time.

Connecting with the Claude.ai connector

To connect Adverity to Claude with MCP, follow these steps:

Open Settings in Claude.ai and navigate to Connectors.
Click Add custom connector.
Configure the connection:

Name
Enter the name for your Adverity connection.

Remote MCP server URL
Enter the URL of the Adverity MCP server: https://mcp.eu.adverity.com/mcp.
Click Add. You will be redirected to the Adverity MCP sign-in page.
Sign in to Adverity. For more information, see Signing in to Adverity MCP.

As a result, your Adverity instance is connected to Claude.

Connecting with Claude Code

Claude Code connects to MCP servers from the terminal. This method does not use the Connectors settings, so it also works when your workspace administrator has restricted custom connectors.

To connect Adverity to Claude Code, follow these steps:

Add the Adverity MCP server:

claude mcp add --transport http adverity-mcp https://mcp.eu.adverity.com/mcp

Start the sign-in:
```
claude mcp login adverity-mcp
```
On the Adverity MCP sign-in page, sign in to Adverity. For more information, see Signing in to Adverity MCP.
Confirm the connection:
```
claude mcp list
```
The server is listed as adverity-mcp with the status Connected.

As a result, your Adverity instance is connected to Claude Code.

Connecting multiple Adverity instances

Claude Code can connect to several Adverity instances at the same time, for example a test instance and a production instance. Add the MCP server once per instance, giving each one a distinct name.

The MCP server URL is the same for every instance. What connects an entry to a particular Adverity instance is the instance URL you enter when you sign in, which is stored in that entry’s own credentials. Each entry keeps its own sign-in, so connecting a second instance does not affect the first.

The name you choose keeps the two sets of tools apart, so that Claude does not confuse the tools of one instance with those of another. Choose names that make the instance obvious, such as adverity-mcp-test and adverity-mcp-production.

To connect an additional Adverity instance, follow these steps:

Add the MCP server under a name that identifies the instance:

claude mcp add --transport http adverity-mcp-test https://mcp.eu.adverity.com/mcp

Sign in to that entry:
```
claude mcp login adverity-mcp-test
```
On the Adverity MCP sign-in page, enter the URL of the instance you want this entry to use, and sign in to that instance. For more information, see Signing in to Adverity MCP.
Repeat these steps for each additional instance, using a different name each time.

As a result, Claude Code is connected to each Adverity instance separately, and claude mcp list shows every entry as Connected.

Managing Claude tool permissions

In Claude, you can control how the assistant executes specific MCP tools:

Always allow (recommended): Claude can use the tool without user confirmation, which allows it to work efficiently without interrupting for permission on every query.
Needs approval: Claude must ask for permission each time, providing more control but slower execution.
Blocked: Prevents Claude from using specific tools entirely.

Connecting to ChatGPT

For ChatGPT, an administrator needs to create the connector first, then individual users can link their accounts.

Setting up the connector (administrator)

To set up the Adverity MCP connector in your ChatGPT workspace, follow these steps:

In ChatGPT, click your workspace name in the bottom-left corner and select Workspace settings.
Go to the Connectors tab.
Click Create.
A pop-up will appear warning about developer mode. Click Enable developer mode.
Configure the connection:

Name
Enter the name for your Adverity connection.

MCP Server URL
Enter the URL of the Adverity MCP server: https://mcp.eu.adverity.com/mcp.

Authentication
Leave this setting empty.
Check the box I understand and want to continue.
Click Create. You will be redirected to the Adverity MCP sign-in page.
Sign in to Adverity. For more information, see Signing in to Adverity MCP.
After signing in, you will be redirected back to the ChatGPT Connectors settings page.
Your new connector will be listed under Drafts. Click Publish.
Acknowledge the potential risks in the pop-up.
Click Publish again.

As a result, the Adverity MCP connection is active and available for all users in your workspace.

Connecting your personal account (user)

Once your administrator has set up the connector, follow these steps to link your personal account:

In ChatGPT, click your name in the bottom-left corner and select Settings.
Go to the Apps & Connectors tab.
Click Adverity in the list of available connectors.
Click Connect.
A pop-up will appear. Click Continue to Adverity.
You will be redirected to the Adverity MCP sign-in page. Sign in to Adverity. For more information, see Signing in to Adverity MCP.
After a successful sign-in, you will be redirected back to your ChatGPT settings. The button will now say Disconnect, confirming the connection is active.

As a result, your Adverity instance is connected to ChatGPT.

Connecting to other tools

MCP is designed for broader agentic workflows beyond chat interfaces. You can integrate Adverity MCP with automation platforms and custom applications using the MCP server URL: https://mcp.eu.adverity.com/mcp.

n8n and Make: Build automated workflows where an AI agent checks if datastreams are successful before triggering the next step in a marketing report.
Custom clients: Build your own MCP clients to allow different internal AI agents to collaborate using Adverity’s harmonized data as a central hub.

Testing your connection

After connecting, verify the setup with the following prompts. Each should return a meaningful response from your Adverity instance:

Show me all my datastreams — returns a list of datastreams from your instance.
Which authorizations are expiring soon? — returns authorization status with expiry information.
Download my latest Google Ads data — returns a CSV data extract from your most recent Google Ads datastream.

Managing your connection

A connection made with Sign in with Adverity uses a Connect API key that Adverity creates for you. Keep the following in mind:

The key is named mcp: followed by the name the AI tool gives itself.
To disconnect the AI tool, revoke the key on the Connect API keys page. If you cannot view Connect API keys, ask an Administrator in the root workspace to revoke it. Removing the connector in your AI tool does not revoke the key. For more information, see Revoking a Connect API key.
The key expires after 30 days. To keep using the AI tool, sign in again.
If you sign in again with the same AI tool and the same workspace, the new key replaces the previous one.
Each key counts toward your limit of 10 personal Connect API keys with an expiry date, together with the keys you create yourself. If you reach the limit, the sign-in fails with an error. Revoke a key you no longer use, or ask an Administrator in the root workspace.

Whichever way you connect, if you do not use the AI tool for some time, it may ask you to sign in again.

What’s next?

Using Adverity MCP