Cube iconCube

Open-source semantic layer with Cube Cloud for agentic analytics: AI data agents, an MCP server and governed metrics over your warehouse

21.2k+ GitHub Stars
$40 Per Developer/mo (Starter)
MCP Server on Free Tier

Overview

Cube is a semantic layer: it sits on top of your data warehouse or lakehouse and defines metrics, joins and access rules once, so dashboards, embedded analytics, APIs and AI agents all get the same governed numbers. Cube Core is open source (Apache-2.0 server, MIT client libraries) and Cube Cloud is the managed platform. The AI feature that puts it on this list is agentic analytics: in June 2025 Cube launched D3, a platform with AI Data Analyst and AI Data Engineer agents built on the semantic layer, and today it sells "Agentic BI" where agents request governed data through an MCP server and APIs, plus natural-language querying, AI-assisted workbooks and access from AI assistants, coding tools and Slack. The MCP server, agents and analytics chat are included even on the free tier. Cube is not a warehouse itself; we list it here because it is the query layer over warehouses and lakehouses such as Apache Iceberg. Cube raised a $25M Series B in 2024 with investors including Databricks Ventures.

The Verdict

Who Should Use Cube?

Best For

  • Data teams that want one set of metric definitions for BI and AI
  • Giving AI agents governed, permission-aware access to data
  • SaaS teams adding multi-tenant embedded analytics

Not Ideal For

  • Teams looking for a storage engine or warehouse
  • Small teams with a single BI tool and few metrics

What's Great

  • Easy to use and customize, per G2 reviewers
  • One shared semantic layer across tools
  • Managed cloud takes build work off the team

Watch Out For

  • Reviewers mention missing features and bugs
  • Documentation can be confusing
  • Small number of G2 reviews so far

Pricing

View all features & details

Key Features

  • Universal semantic layer: metrics, joins, access control
  • Agentic BI with AI data agents (D3, June 2025)
  • MCP server for Claude, ChatGPT, Cursor and Slack agents
  • Natural-language queries and AI-assisted workbooks
  • Embedded analytics from iframes to full data APIs

Company

  • Cube Dev, co-founded by Artyom Keydunov (CEO) and Pavel Tiunov
  • $25M Series B in 2024; investors include Databricks Ventures and Decibel
  • D3 agentic analytics launched June 2025

Real-World Usage

AI Use

  • Agents query governed metrics over MCP and APIs
  • Same definitions shared by BI, embedded apps and chat

Community

  • Open-source Cube Core, repo pushed daily
  • Customer case studies listed on cube.dev

How It Compares

Feature Cube dbt
Main job Serve governed metrics to BI, apps and agents Transform data in the warehouse
Agent access MCP server, AI data agents MCP server, Agent Skills
Open source Cube Core (Apache-2.0 and MIT) Engine (Apache-2.0)
Paid from $40 per developer a month See vendor

Sources: Cube · Pricing · D3 (Jun 2025) · MCP server · TechCrunch (Jun 2024) · GitHub · see also dbt

Is this your tool? Add a badge