The Attentive MCP server connects your AI assistant to Attentive using your authenticated account. This article covers how access, data, and logging work, so your IT or security team can evaluate the connection before you connect.
What data can the MCP access? Only what the connected Attentive user could already see and do in the Attentive UI — access is scoped to that user's existing permissions and companies. The available tools determine what information your AI application can retrieve. The list_segment_subscribers tool can return subscriber phone numbers, email addresses, names, and signup dates, including in read-only sessions. Approval prompts depend on your AI application and its settings.
What does Attentive log? Attentive logs operational security and diagnostic information for MCP usage, including authentication and session lifecycle events, write-tool metadata, and failures with error context. Normal successful tool outputs are not written to application logs by the tool registry.
How does authentication work? You authenticate through Attentive's standard login using OAuth — your AI client opens an Attentive sign-in page in your browser, and after you log in, the client receives an access token. You never enter an Attentive password or API key directly into your AI client's configuration.
Who can connect, and what can admins see? Connect using an Attentive account with access to the companies and features you want to work with. Your organization’s AI workspace policies may require an administrator to enable the connector first.
How do we turn it off? An individual disconnects by removing the Attentive connector from their AI client. Attentive can also turn off access for your company as a whole.
Can the assistant be wrong? What do we do about it? Yes — Attentive does not guarantee the quality of your third-party assistant's outputs or that they match your intent. Custom MCP connections default to read-only mode. Native connectors manage their connection configuration and may include write tools. Approval prompts depend on the AI application and its settings. Review anything before approving it, especially higher-risk actions like scheduling or launching a send, and review the terms and settings for any assistant you connect.
How is personal data handled? Subscriber contact information can be returned in read-only sessions. Read-only access does not mean that all returned data is non-sensitive. Once that data is returned, though, it passes through to your connected AI client, and from there is subject to that provider's own data processing and retention practices.
Does my AI provider train on Attentive data? Tool inputs and results are passed to whichever AI client you've connected (Claude, ChatGPT, etc.) and may be processed or retained according to that provider's own policies — Attentive does not set those terms. Review your AI provider's data retention and training policies directly before sending personal or sensitive data through the connection.