Cube: The Semantic Layer for Agentic Analytics

You’ve got a BI tool, an AI assistant, an embedded analytics layer for customers, and probably a spreadsheet someone insists is the real source of truth. Every one of them is pulling from a slightly different definition of revenue, win rate, or churn. Your analysts are spending more time reconciling numbers than surfacing insights. And when someone asks the AI a business question, the answer is either confidently wrong or impossible to verify.

This isn’t a data quality problem — it’s an architecture problem. When analytics surfaces operate in silos, there’s no shared foundation to enforce consistent logic. Governance becomes a manual exercise. Trust erodes. And as you layer in generative AI, the gap between what the model says and what the data actually shows gets wider, not narrower.

Cube Is the Semantic Foundation That Ties It All Together

Cube is the agentic analytics platform built on a universal semantic layer — a single governed model that grounds AI chat, workbooks, dashboards, and embedded analytics in the same business logic, so every answer ties back to the same numbers.

Cube’s core value is deceptively simple: define a metric once, and every downstream tool — AI, BI, embedded analytics, spreadsheets — uses that definition. Dr. Jun Huang, Global Head of Data Science at Alcon, described this directly: without Cube, analysts were writing 20 different queries for a single core business metric. With Cube, it’s defined once in the data model, and every tool downstream inherits both the definition and the calculation logic.

That consistency has compounding effects. When your AI assistant is grounded in the same semantic model as your dashboards, the answers it generates are traceable and auditable — not hallucinated from raw schema. When you’re shipping embedded analytics to customers, multi-tenancy and row-level governance flow from the same model rather than being bolted on at the query level. And when your product, engineering, and business teams are working from the same metric definitions, the time spent resolving conflicting numbers drops sharply.

Cube also integrates with the rest of your stack without forcing you to rebuild it. Webflow runs Cube Cloud alongside ClickHouse for fast query execution while maintaining the abstraction that keeps different teams from needing to understand database-specific complexity. Brex evaluated dbt’s Semantic Layer and LookML before choosing Cube — specifically because the Semantic Layer is what makes AI useful at scale.

For SaaS companies shipping analytics to customers, Cube’s embedded layer is built for that use case from the ground up: custom branding, your agent name, your color scheme — Cube’s surfaces disappear into your product while governance flows through to each customer’s permissions.

Cube Features

Cube’s features cover the full analytics surface area:

• Analytics Chat API: Build a fully custom AI analytics experience for internal or customer-facing use. Agent-to-agent-capable via the MCP (model context protocol), so it integrates into broader agentic workflows.
• Connected BI: Keeps your business intelligence tools in sync with the semantic layer, so BI outputs stay consistent with every other analytics surface.
• Core Data APIs: Maximum control at the data layer with no prescribed UI — build any interface on top using Cube’s APIs directly.
• Creator Mode: Full workbook and dashboard creation embedded inside your product, giving end customers the ability to build their own reports without leaving your application.
• Cube D3 (Agentic AI): AI data agents that automate reporting, deliver consistent semantics across queries, and surface an explanation for every decision — making AI-driven analytics transparent and auditable.
• Embedded iframes: The fastest drop-in path for embedded analytics — Analytics Chat and Dashboard iframes that can be deployed directly into your product.
• LLM & AI Semantic Layer: Combines Cube’s semantic model with large language models (LLMs) to increase the accuracy of generative AI outputs. Works natively with Anthropic Claude models and supports bring-your-own-LLMs from other top providers. Outputs are traceable back to governed, auditable data.
• Modern Cloud OLAP: Bridges your modern data stack and spreadsheet consumers, so analysts working in Excel or Google Sheets pull from the same governed model as every other tool.
• Semantic Layer (Universal): The core of the platform — a single governed model that defines business logic, metric calculations, and access controls once, then serves every analytics surface consistently.
• Workbooks: Governed report building that gives data teams and business users a structured interface for ad hoc analysis, all grounded in the semantic model.

Cube was recognized in the 2026 Gartner Market Guide for Agentic Analytics and has an active open-source community with nearly 20,000 GitHub stars and over 13,000 members in its Slack community.

Cube becomes our single source of truth for metric definitions and powers everything from customer-facing dashboards to AI-driven quarterly business reviews. CSMs gain back dozens of hours each quarter, enabled by Cube’s semantic layer and agentic analytics.

Anthony Cronander, Senior Analytics Engineer, Drata

Taken together, Cube gives data teams a way to stop maintaining parallel definitions across disconnected tools and start operating from a single governed foundation. Whether you’re delivering internal BI, building AI-powered analytics into your product, or trying to make your LLM outputs trustworthy enough to act on, the semantic layer is what makes that possible at scale.

One Foundation, Every Surface

If your team is managing separate definitions for the same metrics across BI, AI, and embedded analytics, the problem compounds every time you add a new tool or a new customer-facing feature. Cube solves that at the architectural level—not with a workaround. With a semantic layer that governs everything from AI answers to embedded dashboards, you get consistent outputs your team can stand behind, and your customers can trust.

Try Cube for Free!

Frequently Asked Questions