Hubble’s Model Context Protocol (MCP) integration allows supported AI clients, such as ChatGPT or Claude, to securely access and interact with research data in your Hubble workspace. The actions available to each user depend on their Hubble role and permissions.
What is MCP?
MCP is an open standard that allows AI applications to connect with external tools and data sources. With Hubble’s MCP integration, users can ask an AI client to find studies, review research data, analyze responses, and perform supported research workflows without manually copying information between tools.
What MCP tools and actions does Hubble expose?
Hubble MCP supports tools across several areas of the research workflow, including:
- Finding and reviewing studies
- Reading study details, tasks, reports, recordings, and collaborators
- Reviewing participants and individual responses
- Analyzing study results and research data
- Managing recording notes and research evidence
- Creating and updating studies
- Managing moderated-study scheduling and bookings
- Creating and managing recruitment drafts
- Publishing supported studies and recruitment projects
The exact tools available may change as Hubble expands its MCP capabilities. Tool availability also depends on the user’s workspace role, study permissions, and the MCP client being used.
Actions that can affect a live study, participant experience, scheduling, or recruitment may require explicit user confirmation before they are completed.
Does Hubble use customer data to train AI models?
No, Hubble does not grant an MCP client unrestricted ownership or use of customer research data.
Can viewers modify data through MCP?
No. Viewers have read-only access.
They can use MCP to retrieve and review data they already have permission to access, but they cannot create, edit, publish, or delete studies, projects, responses, notes, recruitment settings, or other workspace data.
MCP does not grant users additional permissions beyond those assigned to them in Hubble.
Does MCP give an AI client access to the entire workspace?
No. Access is limited by the permissions of the Hubble user who connects the MCP client.
The AI client can only access the workspaces, studies, and data that the connected user is authorized to access. Connecting Hubble through MCP does not bypass Hubble’s existing role-based access controls.
Is data sent to the MCP client?
When a user requests information through an MCP-connected AI client, the data required to complete that request may be returned to the client.
How that client stores, processes, or retains the returned data is governed by the client provider’s terms, privacy policy, and enterprise settings.
What audit logs exist for MCP and AI actions?
Hubble tracks the source of certain actions performed within the platform. For example, recording notes can identify whether they were created through the Hubble application, generated by AI, or added through MCP. Notes created via AI displays a Generated by AI indicator.
Please contact Hubble Support for details about the logs available for your organization, including any security, access, or administrative audit requirements.
Can MCP make changes without the user knowing?
MCP tools are designed with permission and confirmation controls. Sensitive actions, such as publishing a study, changing participant scheduling, or editing research data, require the user to explicitly approve the action.
Please review AI-generated changes before confirming them, particularly when the action affects a live study or participant experience.
Can workspace administrators disable or restrict MCP?
MCP availability and access may depend on your organization’s workspace settings and Hubble plan. Contact your workspace administrator or Hubble Support to confirm whether MCP is enabled and which users are permitted to connect it.
Where can I get more information?
For questions about MCP permissions, data handling, or security information, please contact Hubble Support or your Hubble account representative.
To learn more about MCP and governance, you can also refer to our MCP guide.