Kubit

What Kubit MCP server does

The Kubit MCP server exposes Kubit analytics functions to an MCP-compatible AI client. It is designed for conversational exploration of warehouse-backed data, including event activity, properties, dimensions, reports, and dashboards. Users can ask focused questions about conversion, retention, engagement, or other analyses without manually constructing query syntax.

The service is useful when an analyst or product team needs to discover available data, reuse existing analyses, generate a new report, or retrieve detailed records for work outside Kubit. The results remain governed by the user’s existing access in Kubit.

How it works

A typical session starts by initializing the user context. The assistant can then inspect available schemas and examine the events, properties, and dimensions within them. Before generating a new analysis, it can search the organization for an existing report or dashboard. If no suitable analysis exists, it can create and execute a report from the user’s request, then export raw results as CSV when needed.

Specific prompts produce better results. Include relevant date ranges, event names, segments, schema names, and the business question being investigated. For complex analysis, split the work into smaller requests rather than asking for several unrelated metrics at once.

Setup and configuration

The Kubit MCP server is accessed at https://mcp.kubit.ai/mcp. Add that endpoint to an MCP-compatible client and complete the OAuth 2.0 sign-in flow when prompted. A Kubit account with access to an organization and at least one permitted schema is required.

No additional credentials or environment variables are documented for this remote setup. The client configuration process varies by product, so follow the selected client’s instructions for adding a remote MCP server.

Tools and capabilities

The Kubit MCP server provides five tools:

• getUserContext initializes the session and returns schemas available to the user.
• getSchema describes events, properties, and dimensions in a selected schema.
• createReport builds and runs an analytical query.
• getRawData exports CSV data from an existing report.
• searchKubit locates reports and dashboards across the organization.

A practical sequence is to initialize the session, inspect the relevant schema, search for reusable content, create a report if necessary, and export only when detailed data is required elsewhere.

Limitations and notes

The Kubit MCP server is identified as being in beta, so users may encounter bugs, performance problems, or rate limits. Report creation depends on the referenced schema and events being available; inspecting the schema first can help resolve failed or ambiguous requests.

Access is restricted to schemas authorized for the signed-in Kubit user. The documentation also notes that third-party AI models process query data, so organizations should review their policies before enabling AI-assisted analysis. Narrow date ranges and selective exports can help keep requests focused and reduce unnecessary workload.