The AdRoll MCP server lets AI assistants, agents, and MCP-compatible clients securely connect to supported AdRoll data and tools.
With the MCP server, your team can bring AdRoll into the AI and automation workflows you already use. You can retrieve campaign performance, create campaign drafts, manage supported campaign workflows, and explore ABM account intelligence without manually exporting data or switching between systems.
This capability is available in open beta and is built using Model Context Protocol (MCP), an open standard for connecting AI systems to external platforms, tools, and services.
What is Model Context Protocol (MCP)?
Model Context Protocol (MCP) is an open standard that gives AI systems a consistent way to connect to external data sources, tools, and services.
Instead of building a custom integration for every AI client and business platform, MCP provides a shared connection layer. This allows MCP-compatible clients to retrieve structured data, invoke supported tools, and use live platform context when responding to a request or completing a workflow.
What is the AdRoll MCP Server?
The AdRoll MCP server is AdRoll’s MCP-compatible connection layer. It exposes supported AdRoll data and tools to external AI systems and MCP-compatible clients through a secure, permissioned interface.
Connected clients can use the MCP server to retrieve AdRoll data or invoke supported AdRoll tools based on the authenticated user’s permissions. For example, a connected client may retrieve campaign performance data, help create a campaign draft, surface ABM account intelligence, or support selected campaign management workflows.
The MCP server is different from the AdRoll AI Assistant, which operates directly inside the AdRoll platform. The MCP server extends AdRoll into external clients such as ChatGPT, Claude, Cursor, n8n, and Microsoft Copilot Studio.
If you want to learn more about using AI features within AdRoll itself, see Getting Started with the AdRoll AI Assistant.
What You Can Do
The AdRoll MCP server supports several workflows across AdRoll ABM.
Reporting and Performance Analysis
Use natural language to retrieve performance data, analyze trends, compare time periods, and summarize results.
Example prompts:
- “Show performance for the last 30 days.”
- “Which campaigns had the biggest week-over-week change in conversions?”
- “What changed in performance this week?”
Campaign Creation and Management
Create campaign drafts, review recommendations, and manage supported campaign workflows from MCP-compatible clients.
New campaigns created through MCP will be draft-first, meaning they’ll be staged for review before they go live. This helps keep marketers in control while making it easier to move from insight to action.
Example prompts:
- “Create a retargeting campaign draft for recent site visitors.”
- “Draft a campaign targeting high-intent accounts.”
- “Help me identify campaigns where budget or pacing should be reviewed.”
Account Exploration and Prioritization for ABM
For AdRoll ABM users, MCP can help teams explore account intelligence, review engagement signals, and prioritize accounts for sales and marketing follow-up.
Example prompts:
- “Which accounts have shown the highest intent activity recently?”
- “Which accounts should our SDR team prioritize?”
- “Are there any high-intent accounts we haven’t targeted yet?”
- “Help me draft outreach for top engaged accounts.”
How the MCP Connection Works
The MCP connection includes two main components.
MCP-Compatible Client
The MCP-compatible client is the application, assistant, agent platform, development environment, or workflow tool that connects to the AdRoll MCP server.
Examples include ChatGPT, Claude, Cursor, n8n, and Microsoft Copilot Studio.
The client sends requests to the MCP server when AdRoll data or supported AdRoll tools are needed. Depending on the client, the returned information may be used to answer a question, continue an automation, generate a summary, or support a workflow.
AdRoll MCP Server
The AdRoll MCP server acts as the secure bridge between the connected client and the AdRoll platform. It exposes supported AdRoll data and tools through a structured, permissioned interface.
For example, if a user asks, “How did my retargeting campaigns perform last week?” the connected client can request the relevant campaign data through the AdRoll MCP server. The server retrieves the data from AdRoll and returns it to the client so it can summarize the results or continue the workflow.
Getting Started
To use the AdRoll MCP server, configure AdRoll as a remote MCP server in your MCP-compatible client.
You will need:
- An active AdRoll account
- Access to an MCP-compatible client
- Appropriate user permissions within AdRoll
- The ability to configure a custom connector, custom app, or remote MCP server connection in your selected client
Some clients may require paid, team, enterprise, developer, or admin-level access to configure MCP connections. Setup requirements vary by platform.
AdRoll MCP Server Endpoint
Use the following endpoint when configuring your MCP connection:
https://services.adroll.com/mcp
Connecting Your MCP-Compatible Client
Setup steps vary depending on the client or platform you are using. In many cases, you will need to configure AdRoll as a custom connector, custom app, or remote MCP server connection.
Use the AdRoll MCP server endpoint above when prompted for the MCP server URL.
Claude
To connect AdRoll to Claude, your team may need to configure a custom connector using a remote MCP server.
For step-by-step instructions, see Anthropic's guide on Getting Started with Custom Connectors Using Remote MCP.
When configuring the connector:
- Use the AdRoll MCP server endpoint
- Follow Claude’s instructions for custom connectors using remote MCP
- If prompted for an OAuth client ID or secret, those fields can be left blank unless otherwise directed by your organization
ChatGPT
To connect AdRoll to ChatGPT, your team may need to configure a custom MCP app using Developer Mode.
For step-by-step instructions, see OpenAI's guide on Developer mode and MCP apps in ChatGPT.
When configuring the app:
- Use the AdRoll MCP server endpoint
- Follow ChatGPT’s instructions for Developer Mode and MCP apps
- If prompted for an OAuth client ID or secret, those fields can be left blank unless otherwise directed by your organization
Cursor
Cursor supports MCP server connections for development workflows.
For step-by-step instructions, see the Cursor's guide on Customizing Model Context Protocol (MCP).
When configuring Cursor:
- Use the AdRoll MCP server endpoint
- Follow Cursor’s instructions for installing or configuring MCP servers
- Confirm that your setup supports remote MCP server connections
Microsoft Copilot Studio
Microsoft Copilot Studio supports MCP integrations for extending agents and workflows.
For step-by-step instructions, see the Microsoft Copilot Studio's guide on Extend your agent with Model Context Protocol.
When configuring Microsoft Copilot Studio:
- Use the AdRoll MCP server endpoint
- Follow Microsoft’s instructions for adding MCP-based actions to agents
- Confirm that your environment and permissions allow external MCP connections
n8n
n8n supports MCP integrations for connecting AI workflows and automations to external systems.
For step-by-step instructions, see the n8n's guide on the MCP Client Tool node.
When configuring n8n:
- Use the AdRoll MCP server endpoint
- Follow n8n’s instructions for using the MCP tool node
- Confirm that your workflow is configured to authenticate with AdRoll before retrieving account data or invoking supported AdRoll tools
Authentication and Permissions
AdRoll uses OAuth 2.1 to securely connect your MCP-compatible client to your AdRoll account.
When you connect through a client, you will be redirected to log in to AdRoll and authorize access. Once authorized, the AdRoll MCP server can respond to requests from that client based on your existing AdRoll permissions.
The AdRoll MCP server also supports Dynamic Client Registration (DCR). DCR helps simplify setup by allowing compatible clients to register dynamically during the connection process, rather than requiring a manually issued client ID and secret in advance.
Access through MCP follows the same permissions model as your AdRoll account:
- The connected client can only access data and tools available to your authenticated AdRoll user account.
- Users can only access the accounts, campaigns, features, and data they already have permission to view in AdRoll.
- MCP does not expand access or override existing permissions.
- Data is retrieved only when needed to respond to a query or complete a supported workflow.
For example, sales teams exploring ABM accounts must already have access to AdRoll ABM. If a user does not have access to certain accounts, features, or data in AdRoll, they will not be able to access them through MCP.
Security
The AdRoll MCP server is designed to provide secure, permissioned access to supported AdRoll data and tools.
Key security principles include:
- User-scoped access: MCP mirrors the permissions of the authenticated AdRoll user.
- OAuth-based authorization: Users authenticate through AdRoll rather than sharing credentials with the connected client.
- No credential storage in the client: Users authorize access through OAuth instead of manually entering AdRoll credentials into a third-party workflow.
- Controlled access: Requests are scoped to supported AdRoll data and tools.
Data usage by the AI provider depends on the client or platform being used. We recommend reviewing the privacy, security, and enterprise data policies for the AI provider or workflow platform used by your organization.
Troubleshooting
Authentication failed
Make sure you are using the correct AdRoll credentials and that your user role has the appropriate permissions for the account you are connecting.
If your organization uses team or admin-managed clients, confirm that your IT or operations team has allowed the custom connector, custom app, or remote MCP server configuration.
Missing data
If the connected client cannot find a specific campaign, account, or report, verify that you signed in with the correct AdRoll account during the authorization process.
You can also ask the connected AI assistant: “Which AdRoll account am I connected to?”
If you manage multiple AdRoll accounts under a single organization, you can access those accounts through the MCP-compatible client based on your existing permissions. When asking questions, specify which account you’re referring to so the client can retrieve the right data.
If you’re unsure whether your accounts are linked under the same organization, or you’re having trouble accessing an account you expect to see, contact our support team at support@adroll.com.
MCP connection is not recognized
If the connected client does not recognize the MCP server connection:
- Confirm the MCP server endpoint was entered correctly
- Verify that the client supports remote MCP server connections
- Confirm the connection was added as an MCP server, custom connector, or MCP app, depending on the platform
- Check whether your client requires a specific plan, admin permission, or developer mode to use MCP
Feedback
The AdRoll MCP server is currently in open beta. Capabilities will continue to evolve as we expand supported workflows, clients, and use cases.
Please provide feedback through this form.
Frequently Asked Questions
What MCP-compatible clients and platforms are supported?
The AdRoll MCP server is designed to work with MCP-compatible AI assistants, agents, development environments, and workflow platforms.
Common examples include:
- ChatGPT
- Claude
- Cursor
- Microsoft Copilot Studio
- n8n
Compatibility and setup requirements may vary by platform. We are continuing to evaluate supported workflows across the MCP ecosystem.
Do I need a specific package to use MCP?
AdRoll access
During the beta, the AdRoll MCP server is supported for all AdRoll packages. As MCP capabilities expand, certain features or workflows may become available only with specific AdRoll packages or permission levels.
MCP client access
Some MCP-compatible clients may require a paid, team, enterprise, developer, or admin-enabled plan to configure custom connectors, custom apps, or remote MCP server connections. Requirements vary by client.
Review the setup documentation for the MCP-compatible client you plan to use before getting started.
Does the AI provider use my AdRoll data to train its models?
Data usage depends on the AI provider or platform being used. For example, ChatGPT, Claude, Cursor, Microsoft Copilot Studio, and n8n each have their own privacy, security, and data usage policies.
Many organizations require employees to use team, enterprise, or admin-managed versions of these platforms because they typically provide stronger controls over how data is stored, retained, and used for model training. These plans may also offer additional security, compliance, and administrative controls that are not available in individual or free accounts.
We recommend reviewing the privacy and enterprise data policies for the provider or platform used by your organization before connecting AdRoll through MCP.
What data and tools can the MCP server access?
The MCP server can expose supported AdRoll data and tools available to your authenticated AdRoll user account. This may include campaign performance data, reporting insights, attribution information, audience information, account data, ABM account intelligence, campaign draft creation, and selected campaign management workflows.
Access depends on your existing AdRoll permissions.
Can MCP make changes to my campaigns?
Yes. The MCP server supports selected campaign management actions for users with the appropriate AdRoll permissions.
If your AdRoll user account has permission to make campaign changes, supported actions may be available through your connected MCP-compatible client. This can include campaign management workflows such as reviewing campaign settings, making supported updates, or creating new campaigns.
New campaigns created through MCP are draft-first, meaning they are staged for review before going live. This helps ensure marketers remain in control of campaign decisions and spend.
Supported write capabilities may vary during beta and will continue to expand over time.
Is there a limit to how many queries I can run?
The MCP server follows AdRoll’s standard API rate limits. High-volume usage may result in temporary throttling.
Is this the same as the AdRoll AI Assistant?
No. The AdRoll AI Assistant operates directly inside the AdRoll platform and supports in-platform AI experiences.
The AdRoll MCP server extends supported AdRoll data and tools into external MCP-compatible clients, allowing systems such as ChatGPT, Claude, Cursor, n8n, and Microsoft Copilot Studio to securely connect to AdRoll.
What if my preferred AI platform does not support MCP?
The AdRoll MCP server requires an MCP-compatible client or platform. If your preferred AI platform does not support MCP, you may need to use another supported client or check whether that platform has added MCP support.
We recommend reviewing the documentation for your preferred AI platform to confirm whether it supports remote MCP server connections.