Cube is an open-source, headless semantic layer for BI, embedded analytics, and AI agents. Compare Cube Core vs Cube Cloud pricing, features, and alternatives.
Cube is a headless, API-first semantic layer that lets data teams define metrics, dimensions, and access rules once and serve them consistently to BI tools, embedded analytics, and AI agents. Founded in 2016 by Artyom Keydunov and Pavel Tiunov and headquartered in San Francisco, the company evolved from the earlier Cube.js JavaScript framework into a dedicated semantic-layer platform.
Cube deliberately does not ship its own visualization front end. Instead it focuses on being the single source of governed metric definitions that Tableau, Power BI, internal apps, and AI copilots can all query consistently through SQL, REST, and GraphQL APIs.
Cube Core is open source (Apache 2.0 backend, MIT client libraries) with roughly 20,000 GitHub stars, and connects to most major SQL warehouses and databases while exposing a Postgres-compatible SQL API alongside REST and GraphQL endpoints.
Cube Cloud, the managed commercial product, adds Cube Store caching and pre-aggregations for performance, a semantic layer IDE, embedded dashboards, and agentic analytics support so AI copilots query governed metrics instead of raw, ungoverned tables.
Cube Cloud offers a free-forever tier for small projects, a Starter tier at $40 per developer per month plus usage-based Cube Compute Unit consumption ($99/month minimum), and a Premium tier at $80 per developer per month plus consumption with a $10,000 annual commitment.
Enterprise pricing is custom and adds a 99.99% SLA, single-tenant and bring-your-own-cloud deployment, custom LLM integration, SSO/SAML, and a DAX API for Power BI connectivity; the fully open-source Cube Core remains free to self-host indefinitely.
A semantic layer defines business metrics, dimensions, and access rules once so every tool that queries the data gets consistent results. Cube implements this as a headless platform accessible via SQL, REST, and GraphQL APIs.
Yes. Cube Core, the underlying engine, is open source under Apache 2.0 (backend) and MIT (client) licenses and can be self-hosted for free. Cube Cloud is the company's paid, managed version.
Cube Cloud has a free tier, a Starter tier at $40/developer/month plus usage-based consumption, a Premium tier at $80/developer/month plus consumption with a $10K/year commitment, and custom Enterprise pricing.
Yes. Cube exposes a Postgres-compatible SQL API, so tools like Tableau, Power BI, and Looker Studio can connect to it as if it were a regular database, in addition to REST and GraphQL APIs.
Cube markets itself as an agentic analytics platform, letting AI copilots and LLM-based agents query governed metric definitions instead of raw tables, which reduces the risk of incorrect or hallucinated figures.
Cube's closest competitors include dbt Labs' Semantic Layer (MetricFlow), Looker's LookML modeling layer, AtScale, and Kyligence.