Cube (semantic layer)

What it is

Cube is a semantic layer: metrics, dimensions, joins and access rules are defined once and served through SQL, REST and GraphQL APIs to BI tools, embedded analytics and AI agents. The vendor now brands itself an “agentic analytics platform” (cube.dev, fetched 2026-10-07).

Maker and history

Cube Dev (Cube.dev); the open-source project cube-js/cube started as Cube.js (GitHub repo created 2018-09-16). CEO Artyom Keydunov wrote that he and Pavel Tiunov started Cube in 2019 (Cube blog, 2024-06-06). Funding (CEO posts on cube.dev): USD 6.2M round led by Bain Capital Ventures (2020-09-29); USD 15.5M Series A led by Decibel (2021-07-19); USD 25M completed 2024-06-06, with Databricks joining as a strategic investor and 645 Ventures adding a board observer. Cube is one of the 16 founding members of Open Semantic Interchange announced 2025-09-23 (open-semantic-interchange).

Editions and deployment

• Cube Core: open source; dual licence - Cube Client MIT, Cube Backend Apache 2.0 (README). About 21k GitHub stars; repo active 2026-10-07.
• Cube Cloud: hosted platform (workbooks, dashboards, agents, embedded analytics). Pricing: https://cube.dev (no amounts recorded).
• Self-host or cloud; multi-tenant embedded analytics supported.

Core architecture

Data stays in the warehouse (Snowflake, Databricks, BigQuery and other SQL sources); Cube generates SQL and adds a built-in caching/pre-aggregation engine for sub-second latency. Models are code (YAML/JS). Data model is compatible between Core and Cloud (README).

Role in an enterprise AI rollout

Grounding layer for text-to-SQL and agents: the LLM queries named metrics rather than writing joins over raw tables. See semantic-layers-compared.

AI features as of October 2026 (vendor site; GA status not stated)

MCP server (docs.cube.dev/docs/integrations/mcp-server; connects Claude, ChatGPT, Cursor, custom agents); text-to-SQL grounded in semantic definitions; Analytics Chat; Workbook Agent; Dashboard Agent; 2026 additions: AI in Spreadsheets, Chat Artifacts. Dated launches on cube.dev: Embedded Agentic Analytics (2026-06-11) and MCP Connectors (2026-06-19: the Cube agent can reach Notion, Linear, Sentry and Attio while answers stay grounded in the semantic layer; docs page admin/ai/mcp-connectors). Tier limits not verified.

Integrations

Warehouses above, BI tools through SQL API, dbt models as sources, OSI/Ossie spec (open-semantic-interchange).

Strengths and weaknesses (opinion)

Strengths: open core, headless/API-first, strong embedded-analytics story. Weaknesses: you maintain a second modelling layer next to dbt; best AI features sit in Cube Cloud.

Self-learning

• Docs: https://docs.cube.dev (loads)
• Repo: https://github.com/cube-js/cube

Sources (fetched 2026-10-07)

https://cube.dev ; https://github.com/cube-js/cube ; GitHub API (pushed 2026-10-08, 20,971 stars; licence shown as NOASSERTION because of the dual licence); Snowflake OSI blog (see open-semantic-interchange). Second pass (2026-10-08): https://cube.dev/blog/cube-dev-raises-62m-to-accelerate-cubejs-development ; https://cube.dev/blog/our-series-a ; https://cube.dev/blog/cubes-raises-25-million ; https://cube.dev/changelog/2026-06-19-changelog-mcp-connectors ; https://cube.dev/blog/announcing-embedded-agentic-analytics ; https://cube.dev/sitemap.xml and docs sitemap (discovery).

Open items

Total funding across rounds, exact company founding date (CEO says 2019), certification, per-feature GA status: unverified.