Cube is a semantic layer with a BI front end built for humans and AI agents. Metrics and dimensions are defined once in a semantic model; dashboards, workbooks, analytics chat, embedded analytics and external agents all query that model, so every answer uses the same definitions.
Product Overview
For revenue teams whose numbers live in a warehouse, Cube is one way to make created pipeline mean one thing everywhere. The semantic model holds the metric logic; an MCP server gives Claude, ChatGPT and Cursor governed access to it; a Slack agent answers in thread. Pre-aggregation caching keeps queries fast. Cube also sells embedded, multi-tenant analytics for product companies.
RevOps Jobs-to-Be-Done
- Define revenue metrics once — Model pipeline, bookings and NRR in the semantic layer so every dashboard and agent computes them identically. KPI: No more competing versions of the same KPI.
- Give agents governed data access — Connect Claude or ChatGPT through the MCP server instead of raw SQL access to the warehouse. KPI: Agent answers traceable to approved metric logic.
- Self-serve questions in Slack — Revenue leaders ask in plain language and get answers grounded in the model. KPI: Fewer ad-hoc data requests to RevOps and analysts.
Key Features
- Semantic layer: Central metric and dimension definitions with an IDE and a semantic-model agent.
- MCP server: Available from the Free plan; serves governed data to external agents.
- Analytics chat and workbooks: Natural-language querying and SQL/visual workbooks grounded in the model.
- Dashboards: Conversational dashboard building with pre-aggregated, sub-second performance.
- Embedded analytics: Multi-tenant iframes, APIs and creator mode for SaaS products.
How It Fits Your Stack
Primary system of record: Cloud data warehouse
Key integrations: Snowflake, BigQuery, Databricks, Claude, ChatGPT, Cursor, Slack
Data flows: Queries the warehouse through the semantic model; serves results to dashboards, apps and agents; does not copy source data except cached pre-aggregations.
Security & Compliance
- SSO / SAML: Enterprise plan (Okta, SAML, SCIM)
- Audit logs: Yes
Implementation & Ownership
- Time to first value: Days for a first model on the Free plan; weeks to model a full revenue domain
- Implementation complexity: Medium
- Typical owners: Analytics engineering, Data team, RevOps analyst
Requires revenue data already in a warehouse and someone comfortable writing semantic models.
Pricing & Contracts
- Pricing model: Free tier; per-seat paid plans plus deployment compute
- Indicative range: Free; Starter $40/developer/mo; Premium $80/developer, $40/explorer, $20/viewer per month; Enterprise custom
- Free tier: Yes
- Common add-ons: Dedicated deployment from $0.60/hour, Multi-cluster deployment on Premium
Who It's Best For
Data-mature B2B companies that compute revenue metrics in a warehouse and want BI and agents on one governed model.
Good fit if:
- Revenue data already lands in Snowflake, BigQuery or Databricks
- You want to try an MCP-governed semantic layer for free
- You also need embedded analytics for customers
Probably not ideal if:
- Your reporting lives only in the CRM
- No one on the team can maintain a semantic model
Pros
- Free tier includes the MCP server
- One model serves humans and agents
- Transparent per-seat pricing
Cons
- Needs a warehouse and modeling skills
- Not GTM-specific — no built-in identity resolution across CRM tools
- Full audit logging and SSO only on Enterprise
Often Compared With
- AtScale — AtScale targets large enterprises with Excel/Power BI-heavy estates; Cube is developer-first with a free tier.
- dbt — dbt's Semantic Layer suits teams already defining models in dbt; Cube adds its own BI and caching.
- Looker — Looker's LookML ties the semantic model to Looker; Cube serves any BI tool and agents.
Frequently Asked Questions
What is Cube used for?
Cube is a semantic layer with a BI front end built for humans and AI agents. Metrics and dimensions are defined once in a semantic model; dashboards, workbooks, analytics chat, embedded analytics and external agents all query that model, so every answer uses the same definitions.
How much does Cube cost?
Cube pricing: Free; Starter $40/developer/mo; Premium $80/developer, $40/explorer, $20/viewer per month; Enterprise custom A free tier is available.
What does Cube integrate with?
Cube integrates with Snowflake, BigQuery, Databricks, Claude, ChatGPT, Cursor, Slack.
Who is Cube best for?
Data-mature B2B companies that compute revenue metrics in a warehouse and want BI and agents on one governed model.
What are the best Cube alternatives?
The tools most often compared with Cube are AtScale, dbt, Looker.
