✦ Now with returning players and D1, D7, D30 retention
☰
All featuresAttributionRetentionAnalytics OverviewIntegrationsPricingDocumentationComparisonsFree toolsBlogGlossary
PLATFORM

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 reads approved KPI, time-series, page, source and event reports. list_dashboards lists saved project dashboards. get_profile looks up an explicit application user ID.

Create a dedicated MCP credential

An owner or admin with MFA creates an API key in project integrations. Configure your client for the API’s /mcp endpoint with Authorization: Bearer YOUR_API_KEY. This is a private read credential; never put it in browser code or substitute the Publishable key.

Use an HTTP MCP client that supports authorization headers

Endpoint: https://api.your-domain.com/mcp
Header: Authorization: Bearer YOUR_API_KEY

Ask a reproducible question

Ask for top pages or top events in a specific date range. 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 Publishable key?

No. It uses a separate API key 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.