DOCSEARCHbyAlgolia
Submit Docs

Cube

website·Indexed 39 days ago

https://cube.dev/

An agentic analytics platform built on a semantic layer designed for internal business intelligence and AI-native applications.

Tokens
445
Snippets
0
Records
5
Agent score
74%

What's inside Cube

  1. Overview of Cube and Cube Core

    Cube is an AI-native agentic analytics platform designed for internal business intelligence and embedded analytics. It uses a semantic layer as its foundation to ensure AI-generated answers are trustworthy and governed in production environments.

    Cube consists of two primary components:

    1. Cube Core: An open-source (Apache 2.0) semantic layer that serves as the foundation.
    2. Cube (Commercial Platform): A commercial offering built on Cube Core that adds features such as Analytics Chat, workbooks, dashboards, embedded surfaces, Model Context Protocol (MCP) support, multi-tenancy, and governance.

  2. Expose governed analytics to AI agents via MCP Server

    Cube provides an Analytics MCP (Model Context Protocol) server that allows AI agents to access governed analytics metrics. This architecture ensures that agents operate with proper permissions, traceability, and grounding in the semantic layer, preventing the hallucinations common in raw text-to-SQL approaches.

  3. Implement multi-tenant embedded analytics security

    When implementing multi-tenant embedded analytics, focus on tenant isolation and the flow of the signed security context through requests. Key considerations include preventing cache leaks across different tenants and following a pre-launch security checklist to ensure data isolation.

  4. Connect ChatGPT to a Data Warehouse using Cube MCP

    You can connect ChatGPT Work to governed data warehouse data by using Cube's hosted Model Context Protocol (MCP) server. This setup ensures that AI agents interact with data through semantic definitions, OAuth authentication, and inherited user permissions rather than raw database access.

  5. Add AI-powered analytics to a product

    To ship AI-powered, governed analytics inside a product, you must ground the AI's answers in a semantic layer. This ensures that the AI agent uses governed metrics, permissions, and lineage instead of querying raw tables, which provides traceable and permission-aware answers.