Analytics MCP Gateway

The Volāre Analytics MCP gateway is a single remote MCP server that lets Claude (and any MCP-compatible client) query Volāre’s analytics data in natural language. Instead of wiring up each analytics tool separately, employees connect once and ask questions across every source.

It is hosted as a container on the AWS ops-runner instances and reachable at https://analytics.byvolare.com.

What it connects

The gateway proxies several upstream analytics sources behind one endpoint:

Source	What you can ask about
Google Analytics (GA4)	Traffic, acquisition channels, conversions, realtime users, funnels, custom dimensions/metrics.
Microsoft Clarity	Dashboard metrics (sessions, devices, geography) and session-recording metadata.
Volāre database (read-only)	Business data via curated analytics views — active products per market, offers, lead prices, booking funnel, including client/contact details.

More sources are planned (e.g. Meta Ads, Google Ads). Adding one is a registry change in the gateway — clients don’t need to reconnect.

Endpoints

Each source is exposed on its own path, plus one aggregated endpoint that merges every source’s tools together:

Endpoint	Tools exposed
/mcp/analytics	All sources, namespaced (ga__…, clarity__…, db__…). Recommended for most use.
/mcp/ga	Google Analytics only.
/mcp/clarity	Microsoft Clarity only.
/mcp/db	Volāre database (read-only) only.

On the aggregated endpoint, tool names are prefixed with the source (ga__run_report, db__query, …) so there are no collisions.

Access & authentication

Access is gated by Sign in with Google. Only Google accounts on the byvolare.com or gocimit.com domains are allowed. There is no shared password or API key for interactive use — the OAuth flow handles everything.

The database source can return personal data

The db source exposes booking and offer views that include client and contact names, emails, and phone numbers. Treat anything you retrieve as personal data and handle it accordingly.

How it works

The gateway is a thin proxy. For each enabled source it spawns the upstream MCP server as a child process and forwards tools/list and tools/call requests:

• GA → the official Google analytics-mcp server (authenticated keylessly via GCP Workload Identity Federation).
• Clarity → clarity-mcp-server, authenticated with a Clarity API token.
• DB → mcp-server-postgres pointed at a read-only Postgres role that can only SELECT from the analytics_* views.

A source is enabled only when its credentials are present in the gateway’s environment, so the available endpoints reflect what’s configured.

Querying the database correctly

The db source is generic SQL — it doesn’t ship a tool per question. Claude writes the SQL itself against the curated analytics_* views (e.g. analytics_products_by_market, analytics_offers, analytics_bookings). Those views carry column comments / a data dictionary so the model can map a question like “active products for Spain and their lead price” onto the right columns without knowing the full application schema.

Next steps

• Use it from Claude Desktop