Every equation you find has to pay rent.
Not "1M API calls." Not "vCPU-hours." Equations discovered per month. Each successful /discover counts as one. Predict, drift-check, and share the equation as many times as you want — those are free.
One prototype, one weekend. IP-throttled — no card.
- 50 equations / month
- 1,000 datapoints per fit
- 7-day drift retention
- Community support (Discourse)
- MIT engine (self-hosted stays free forever)
10 sensors, hourly refits = 500 fits/mo. One prosumer rig.
- 500 equations / month
- Unlimited /predict calls
- 30-day drift retention
- Email support · 48hr response
- Custom share URLs
Process line: 200 sensors, 15-min refits = ~5K fits/mo.
- 5,000 equations / month
- Priority GPU queue (no cold-start)
- Custom domain · yourco.formulize.dev
- SLA 99.5% uptime
- Slack support · 24hr response
- Grafana + Node-RED webhooks
Multi-plant / R&D group. Signed MSA + named contact.
- 25,000 equations / month
- Signed SLA + phone support
- SSO · SAML / OIDC
- Multi-user seats · role-based access
- Audit log export
- Onboarding call & office hours
Air-gapped plant, GxP, regulated data. On-prem Docker.
- Unlimited equations
- On-prem Docker · air-gap OK
- Custom cartridges (GxP / SEMI / IEC)
- 24-hour incident response
- Data-residency options (EU / IN / US)
- SOC 2 Lite roadmap · GDPR DPIA
How the meter works
Three simple rules. No hidden overage. If you hit your cap we return HTTP 402 with an upgrade URL — never a surprise invoice.
1 successful fit = 1 equation
A /discover call that returns an equation with dim-check passed and R² ≥ 0.5. Failed fits are free.
Predictions are unmetered
Once you have an equation, hit its endpoint as often as you want. We're not selling inference — we're selling structure discovery.
Rollover, no overage billing
Unused equations roll one month forward. If you exceed cap mid-month we pause — no auto-charged overage until you opt in.
Cheaper than the person you'd hire to write it
The Pro plan handles the workload of a full-time contract data scientist for a mid-size process line — at less than 1% of the cost. And the equation is auditable, not a black-box model.
Hiring a data scientist to fit these models
- Loaded cost: $150–220K/yr FTE or $150/hr contractor × 80 hrs
- 2–6 weeks to ramp on your data pipeline
- Model output: a Jupyter notebook + Slack message
- Rebuild every time the sensor set changes
- Turnover risk · knowledge lives in one head
formulize.dev Pro
- 5,000 equations across every process you meter
- Live in an afternoon · Node-RED node ships today
- Output: closed-form equation + live API endpoint + drift monitor
- Refits automatically on new data
- Documentation and dim-check baked in · reviewable by any engineer