transparent pricing

One engine. Two entry points.

cube profiles your data warehouse once, builds a precomputed answer layer, and serves every question at sub-millisecond latency — with $0 marginal cost per answer after ingest.

Team
Self-serve access for data teams that want answers today, not in Q3. Connect a warehouse, upload files, start asking.
$500 – 2,500
/ month · annual billing · varies by warehouse size
Get started →
  • File upload via /v1/ingest — CSV, JSON, ParquetInstant askspace build on upload
  • Postgres warehouse connectionRead-only connector, credentials never returned to client
  • Natural-language + structured asks against your dataPrecomputed cells, not live queries
  • Full provenance on every answerRow-level trace + Wilson-interval confidence
  • Dedicated API key per org
  • Email support
Enterprise
Enterprise
Dedicated deployment, SLA, per-org tenant isolation, and warehouse connectors at scale. Pricing based on data volume and seat count.
Custom
contact us for a quote · no public list price
Contact sales →
  • Everything in Team, plus:
  • Per-org tenant isolation with dedicated API keysMulti-org support via cube's native org layer
  • Warehouse connectors including Snowflake and BigQueryConnector code ships with cube — exact certification status confirmed at contract time
  • Dedicated deployment — your region, your VPC or ours
  • SLA and uptime commitment
  • Custom ingest schedules and pipeline hooks
  • Slack + dedicated CSM
feature detail

what's in each tier

Feature Team Enterprise
Sub-millisecond answer latency (precomputed) ✓ ✓
File upload ingestion (/v1/ingest) ✓ ✓
Postgres warehouse connector ✓ ✓
Snowflake / BigQuery connectors — ✓
Read-only credential handling (never returned to client) ✓ ✓
Row-level provenance trace on every answer ✓ ✓
Wilson-interval confidence band ✓ ✓
Dedicated API key per org ✓ ✓
Per-org tenant isolation — ✓
Dedicated deployment (VPC/region choice) — ✓
SLA commitment — ✓
Dedicated customer success — ✓
🔒

Credentials never leave your hands.

cube connects read-only. Warehouse credentials are stored server-side and redacted from every API response — no token, URI, or secret is ever returned to the browser or a third party. All answers are computed from prebuilt cells, not live queries that could expose raw credentials in transit.

common questions

before you reach out

How does cube connect to my warehouse?

You supply read-only credentials once. cube profiles the schema, builds a precomputed askspace, then severs the live connection — answers are served from that cached layer, not re-queried each time.

Is there a free trial?

Yes — email us and we'll provision a sandbox with a sample warehouse. No credit card required to evaluate.

What does "sub-millisecond latency" mean in practice?

Answers are precomputed on ingest, not queried live. When a user asks a question, cube looks up the matching cell in the askspace — no warehouse roundtrip, no query planner, no cold cache.

Are Snowflake and BigQuery connectors available today?

Connector code ships with cube. Postgres is proven in production; Snowflake and BigQuery connector availability is confirmed per engagement — ask us about your specific warehouse at contract time.

Can we run cube inside our own VPC?

Yes, on Enterprise. We support customer-hosted deployments with dedicated infrastructure. Team tier runs on cube-managed infrastructure.

How is pricing calculated?

Team is sized by warehouse ingest volume and seat count, billed annually. Enterprise is custom — negotiated on total data volume, org count, SLA tier, and deployment model. No per-answer charges on either tier.