Bring analytics into your AI assistant
Connect an HTTP MCP client to project-scoped read-only tools for analytics reports, saved dashboards, and explicitly identified user profiles.
How to install ↗Three tools with a clear scope
get_analytics_report reads approved KPI, time-series, page, source, event, and Web Vitals reports. list_dashboards lists saved project dashboards. get_user_profile looks up an explicit application user ID.
Create a dedicated MCP credential
An owner or admin with MFA creates an MCP key in project integrations. Configure your client for the API’s /mcp endpoint with Authorization: Bearer YOUR_MCP_KEY. This is a private read credential; never put it in browser code or substitute the public SDK write key.
Use an HTTP MCP client that supports authorization headers
Endpoint: https://api.your-domain.com/mcp
Header: Authorization: Bearer YOUR_MCP_KEYAsk a reproducible question
Ask for top pages in a specific date range or Web Vitals for a route. Report tools return their resolved filters and data rows so the assistant can explain the measurement behind its answer.
Read access, not automated changes
The current server does not create charts, send alerts, or mutate profiles. Revoke a key to remove access. Hosted MCP setup requires the API and authentication database to be configured.
Frequently asked questions
Can MCP modify my workspace?
No. The currently registered tools are read-only.
Does it use the SDK write key?
No. It uses a separate project MCP credential with read access.
Which clients work?
Use a client that supports HTTP MCP and a custom Authorization header. Client setup details vary; OAuth login is not implemented here.
Start with a clear measurement model.
Install the SDK, then explore traffic and product activity with the same definitions.