Observe
Tend reads query metadata, billing exports, and INFORMATION_SCHEMA. Row content never leaves your warehouse.
Private beta · Google CloudBigQuery,
BigQuery,
tuned every five minutes.
A team of AI agents watches your warehouse around the clock. They forecast demand and solve for the cheapest capacity that still meets your business constraints: performance targets, budgets, schedules.
TunesBigQueryDatabrickssoonMicrosoft Fabricsoon
48%
one partner · first six weeks
24 / 7
continuous optimization
Google Cloud
BigQuery reservations
A design partner running $430K a year on BigQuery cut reservation spend 48% in the first six weeks. How we measure savings.
Saved this period
$22,410
↑ 31.2%vs. baseline
Forecast EOM
$128.4K
baseline$185.8K
ActualPre-Tend baselinePolicy ceiling
+200 slots · forecast peak in 14m2m ago
lowered ceiling · demand below p9018m ago
planning re-forecast · 24h peak revised1h ago
Watch the money, not the dashboard.
Cost tools hand you a recommendation and a to-do list. Tend's agents make the change themselves, write down what they did and why, and give you the undo button.
Forecast
Tend models the next demand peak before it arrives, so capacity is provisioned when the rollup runs.
Act
Tend solves for the cheapest capacity that meets your business constraints, then applies the change. Auto-apply stays off until you enable it.
Record
Tend logs the action, the reasoning, and the rollback path. Verify the result, or restore the previous state.
Tend heals itself. When a forecast mis-predicts the workload or it detects sustained SLO violations, Tend reverts to your baseline on its own, keeps verifying, and rejoins the loop once the system converges. You can also disarm and revert to baseline settings at any time.
BigQuery agents
Four agents, one warehouse.
Slot Agent tunes your reservations. BI Engine, Query, and Cost agents each own one surface, inside the same policy limits and rollback controls.
| Agent | Surface | Savings / mo | Actions / week | |
|---|---|---|---|---|
Slot Agent Live BigQuery reservations | edw-prod-reservations | $84.2K | 102 | |
BI Engine Agent Live BI Engine capacity | bi-prod-100gb | $12.5K | 38 | |
Query Agent Live Query analysis | edw-prod.analytics | $18.9K | 21 | |
Cost Agent Live Billing & forecasting | billing-export-v2 | $9.4K | 54 |
What Tend does for a typical workload
Results from three reference reservations.
1,000-slot enterprise reservation
Mixed analytics + ad-hoc SQL
34%
spend reduction · 90d
Slot Agent rebalances burst reservations against forecast demand.
BI Engine + 40 recurring dashboards
Six business units, 200 weekly analysts
2.1×
cache hit rate
BI Engine Agent warms the cache for the dashboards used most.
6-engineer data platform team
Mid-market SaaS · 2 PB total
0
hours / week on warehouse maintenance
Slot, Query, and BI Engine agents handle slot tuning, query review, and BI Engine sizing. The estimate uses engineering time logs from mid-market data platforms with about 6 engineers that spend 4 to 6 hours per week on maintenance.
In motion
A day in motion
You see the agent, the target, the reasoning, the cost impact, and the rollback control for every action. Every entry includes the data that you need to verify the result.
Raised reservation ceiling by 200 slots
applied
Lowered reservation ceiling by 300 slots
applied$142.3 / mo
Planning re-forecast: next 24h peak revised
applied
Rolled back: convergence check failed
rolled-back
Raised ceiling for Monday morning rollup
applied
Confidence hold: nowcast band widened, no change
pending
Flat monthly pricing
Flat price by spend tier. If Tend cuts your spend, your bill doesn't creep up to eat the savings. And if measured savings don't hit 3x our fee in 90 days, the first quarter is free. Current beta pricing covers a single reservation.
Growth
For workloads under $250K annual BigQuery spend
$1,500/ month
BigQuery spend < $250K / yr
- 1 BigQuery project
- Slot Agent
- Read-only by default
- Slack and email alerts
- Community support
Most teams
Platform
For production BigQuery reservations
$5,000/ month
BigQuery spend $250K – $500K / yr
- Up to 5 connected sources (planned after beta)
- All four agents
- Custom policies and approval rules
- Email support
Scale
For large data platforms
Contact us
BigQuery spend $500K – $1M / yr
- Unlimited connected sources (planned after beta)
- Multi-region, multi-project (planned after beta)
- SSO + SCIM (planned after beta)
- All alerting integrations (planned after beta)
- Dedicated solutions engineer
Tier comparison
| Tier | Annual BigQuery spend | Monthly price |
|---|---|---|
| Growth | BigQuery spend < $250K / yr | $1,500 / month |
| Platform | BigQuery spend $250K – $500K / yr | $5,000 / month |
| Scale | BigQuery spend $500K – $1M / yr | Contact us |
| Enterprise | BigQuery spend $1M+ / yr | Custom |
Enterprise · custom
$1M+ spend · sovereign cloud · custom SLA
90-day performance guarantee
For customers with a signed agreement, Clevora refunds the first-quarter subscription fee when measured savings are less than 3 times the monthly Tend fee within 90 days after go-live.
Questions that data leaders ask first
Can Tend make a change that you do not approve?
Tend does not apply a change unless you enable auto-apply for that agent on that warehouse. When you enable auto-apply, Tend applies changes only within the confidence and risk thresholds that you set in policy.
What does Tend cost if you do not see savings?
For customers with a signed agreement, Clevora refunds the first-quarter subscription fee when measured savings are less than 3 times the monthly Tend fee within 90 days after go-live.
Where does your data reside?
Only telemetry leaves your warehouse: query metadata, billing exports, and INFORMATION_SCHEMA. Your data stays in your warehouse.
What is the recovery time if a change is incorrect?
A slot reservation change reverts automatically, typically in under a minute. Tend logs every action, and you can restore the previous state.
How does Tend differ from a recommendation tool?
Most cost tools stop at a recommendation. Tend acts within the scope that you approve and records the result. You measure Tend by actions that Tend applies, not by suggestions.
See your savings forecast
before you commit.
Read access to catalog tables, 30 days of your history, a forecast in your numbers. Then you decide.