Summary

Funnels, retention and event volumes without learning the query syntax.

The workflow the vendor recommends is worth following: get your context, explore the schema, search for an analysis that already exists, and only then create a new report — searching first is both faster and cheaper than generating another query. Two costs to weigh before you connect it: the server is in beta, so bugs and rate limits are expected, and your query data is processed by a third-party model, which is a question for whoever owns that policy at your company.

What it is

Kubit's hosted MCP server for its warehouse-native analytics platform. You connect over OAuth and query schemas, generate reports and export data from a conversation instead of a query builder.

What you get

`getUserContext` — starts the session and lists the schemas you can reach
`getSchema` — the events, properties and dimensions in a schema, in detail
`createReport` — generates and executes an analytical query from your question
`getRawData` — exports CSV from an existing report for analysis elsewhere
`searchKubit` — finds existing reports and dashboards across your org, so you reuse an analysis rather than rebuild it
Your existing Kubit permissions carry over: you only see schemas you were already authorized to view

Requirements

An active Kubit organization account with schema access, and an MCP-capable client. Nothing to install: add https://mcp.kubit.ai/mcp as the server URL and complete the OAuth 2.0 flow when prompted.