Cube is an agentic analytics platform built on a semantic layer, made by Cube Dev, Inc. It centralizes metric definitions, calculations, and security rules in one place. That governed model then feeds every analytics surface, from chat interfaces to dashboards to external AI tools.
The platform targets two common problems. One is fragmented business logic, where teams end up with conflicting versions of the same metric. The other is AI models guessing at database schemas and producing inaccurate answers. Cube addresses both by grounding queries and AI agents in a single semantic model, so a question asked in Slack returns the same answer as one asked through a dashboard.
Teams model data using SQL-first YAML or an AI-powered Semantic Model Agent that connects to major data warehouses. Once modeled, data is exposed through four surfaces: natural-language Analytics Chat, Workbooks for analysts, Dashboards for reporting, and a Model Context Protocol (MCP) server for external AI tools. Pre-aggregations and Cube Store caching are built in to keep query response times fast under concurrent load.
Cube is built on Cube Core, an open-source semantic layer, which lets organizations build customized, white-labeled embedded analytics without re-platforming as usage grows. More than 400 companies use Cube to build embedded analytics into their own products, including Brex, Webflow, and Alcon.
Pricing
Cube offers a free-forever plan with 1 developer seat, 5 workbooks, and 1,000 daily requests. The Starter plan costs $40 per developer per month and includes unlimited workbooks and premium LLMs, but not dbt integration. Premium costs $80 per developer per month, adds dbt integration, Explorer seats ($40/month), Viewer seats ($20/month), embedded dashboards, and a 99.950% uptime SLA. Enterprise pricing is custom and adds SSO, audit logging, and a 99.990% uptime SLA; dedicated compute and caching workers on Starter and Premium are billed hourly on top of seat prices.
* Disclaimer: Please note that pricing information may not be up to date. For the most accurate and current pricing details, refer to the official website.
Key Features
- ✓
Semantic layer defines metrics once using SQL-first YAML
- ✓
Analytics Chat answers plain-English questions with charts and tables
- ✓
Dashboard Agent builds and edits visual dashboards conversationally
- ✓
Workbook Agent drafts and iterates on SQL queries with analysts
- ✓
MCP server connects Claude, ChatGPT, and Cursor to the data model
- ✓
Row-level security prevents cross-tenant data leakage
Use Cases
Centralized Metric Governance
Data engineering teams define core metrics like ARR or active users in one repository. Every downstream BI tool, dashboard, and AI agent then reports the same numbers, which removes conflicting internal reports.
Customer-Facing Embedded Analytics
SaaS product teams embed customized dashboards and AI chat directly into their software. Row-level security keeps each customer’s data isolated, while white-labeling blends the analytics into the host product.
Self-Serve Analytics for Business Users
Non-technical staff in sales or HR ask conversational questions through Slack or Analytics Chat. This reduces the ad-hoc query load on data analysts while keeping answers grounded and accurate.
AI-Augmented Data Exploration
Analysts use Workbooks alongside the AI Workbook Agent to draft, visualize, and refine queries. This speeds up exploration by letting AI handle repetitive SQL work.
Secure External AI Integrations
Developers connect Claude, ChatGPT, or custom agents to corporate data through Cube’s MCP server. Queries respect existing access controls and return structured business context instead of raw tables.
Strengths & Weaknesses
Strengths
Built on open-source Cube Core, which avoids closed-box vendor lock-in.
A centralized semantic layer keeps metric definitions consistent across dashboards, APIs, and AI chat.
Native row-level security supports safe multi-tenant embedded analytics without cross-tenant leakage.
Built-in pre-aggregations and caching address high-latency queries against large data warehouses.
Connects directly to major data ecosystems, including Snowflake, Databricks, BigQuery, and dbt.
Weaknesses
The free plan caps usage at 5 workbooks and 1,000 requests per day.
Non-developer seats and dbt integration are only available on the Premium plan and above.
Enterprise security features like SSO, audit logging, and BYOC require a custom-priced contract.
Dedicated infrastructure and caching workers add hourly compute fees on top of seat pricing.
Who Is This For?
Data engineering and analytics teams: a single place to write and maintain data logic, which reduces ad-hoc requests and conflicting reports.
B2B SaaS product and engineering teams: secure, multi-tenant embedded analytics via APIs and iframes, shipped without re-platforming.
Non-technical business users in marketing, finance, or HR: natural-language chat and Slack integration for answers without writing SQL.
AI application developers: MCP server integrations to connect models like Claude or ChatGPT to governed business data.
Frequently Asked Questions
Is there a free plan for Cube?
Yes. The free-forever plan includes 1 developer seat, up to 5 workbooks, 1,000 requests per day, and basic AI agents.
Does Cube support single sign-on?
Yes, but SSO with SAML 2.0, Okta, and SCIM is exclusive to the custom-priced Enterprise plan.
How does Cube try to prevent AI models from hallucinating data?
Cube grounds AI agents in its governed semantic model rather than raw database tables, so answers reference predefined metrics and logic.
Can I embed Cube’s dashboards into my own product?
Yes. Cube supports embedding via drop-in iframes, a Creator Mode for customer-built dashboards, and REST, GraphQL, and SQL data APIs.
Which plan includes dbt integration?
dbt integration is available starting on the Premium plan ($80 per developer per month) and above; it is not included on Starter.
How does compute billing work for dedicated deployments?
Seat prices are fixed monthly, but dedicated deployments, multi-cluster setups, and caching workers are billed hourly on top of that, starting around $0.15 to $1.20 per hour depending on the resource.
Can I use my own LLM with Cube?
Bring Your Own LLM (BYOLLM) is available to Enterprise customers; other plans use Cube’s supplied premium LLMs.
What payment methods does Cube accept?
Self-serve plans require a credit card and do not support ACH or wire transfers. Order form customers can pay by invoice or wire.
Is Cube open source?
Cube Core, the underlying semantic layer, is open source. The commercial Cube platform adds hosting, AI agents, and enterprise features on top of it.
Cube connects to data warehouses including Snowflake, Databricks, BigQuery, and ClickHouse, with dbt integration on Premium and above. It integrates with Claude, ChatGPT, and Cursor through its MCP server, plus a native Slack agent for in-thread queries. Downstream reporting tools include Excel and Tableau via APIs, and Power BI via a DAX API on the Enterprise plan. Semantic models can also be managed through GitHub for source control.