Cube Dev

Cube is the semantic layer between raw data warehouse tables and the applications that consume metrics — developers define measures and dimensions once in Cube schema, and any consumer (dashboard, data app, AI agent) queries through a consistent API without knowing the underlying SQL. For agents querying business metrics: GET /cubejs-api/v1/load with a Cube query object returns structured metric data without writing SQL. Cube handles caching, query optimization, and pre-aggregations. Self-hostable or Cube Cloud managed. Confidence is docs-derived.

REST API: POST /cubejs-api/v1/load with {query: {measures: ["Orders.count"], dimensions: ["Orders.status"], timeDimensions: [{dimension: "Orders.createdAt", granularity: "month"}], filters: [...]}}. GraphQL API mirrors the REST query model. Meta API: GET /cubejs-api/v1/meta returns available cubes, measures, and dimensions for discovery. SQL API: Cube also exposes a Postgres-compatible SQL endpoint for BI tools. WebSocket API for real-time queries.

API token auth: Authorization: Bearer {api_token} header. Token generated via CUBEJS_API_SECRET. Cube Cloud: workspace tokens. HTTPS enforced. No OAuth2. JWT tokens can include security context for row-level security (e.g., user_id-scoped queries). Self-hosted: configure auth middleware for production security.

Query errors: 400 with detailed error message (e.g., unknown measure name, invalid filter). Cache invalidation: Cube handles query result caching; cache keys include query structure. Pre-aggregation build failures logged in Cube dashboard. Cube Cloud uptime at status.cube.dev. Self-hosted: reliability tracks operator deployment and data source availability.

cube.dev/docs covers schema definition, REST/GraphQL API reference, security model, pre-aggregation guide, and deployment. Getting started: install Cube (npm/Docker), define a schema, first query in under 20 minutes. Cube Cloud for managed hosting. SDKs for React, Vue, and plain JS. Community via Cube Slack.