Every Cruuvo system is tailored. Price reflects the outcomes, capabilities, execution placement, evaluation, usage, and operating responsibility required—not a product tier or level of tailoring.
You do not need to compare model names. Choose how much capability the work needs and whether providers may use customer inputs for model improvement; Cruuvo routes each step within that boundary.
| Strategy | Relative cost | Model selection | Customer inputs used for provider training | Best for |
|---|---|---|---|---|
| Open Weight | $ | Cost-efficient open-weight models | Allowed under provider terms | Routine extraction, classification, summarization, and drafting |
| Smart Router — OpenRecommended · Open data | $$ | Best price-to-quality model for each step | Allowed under provider terms | Workloads where provider training use is acceptable |
| Smart Router — PrivateRecommended · Private data | $$$ | Best price-to-quality private model for each step | Not allowed | Business data that needs a private model boundary |
| Frontier | $$$$ | Highest-capability private models | Not allowed | The hardest reasoning and highest-consequence work |
“Allowed” means a selected provider’s terms may permit customer inputs to be used for model improvement. “Not allowed” means Cruuvo selects eligible services whose terms exclude customer inputs from provider training.
| Feature | Core | Forge | Private Edge |
|---|---|---|---|
| Tailored system design | ✓ | ✓ | ✓ |
| Typed, immutable system versions | ✓ | ✓ | ✓ |
| Policy enforced outside prompts | ✓ | ✓ | ✓ |
| Least-privilege capability bindings | ✓ | ✓ | ✓ |
| Evaluation and safe improvement | ✓ | ✓ | ✓ |
| Per-firm budget cap | ✓ | ✓ | ✓ |
| Audit log per step | ✓ | ✓ | ✓ |
| Human-review gates | ✓ | ✓ | ✓ |
| Versioned releases | ✓ | ✓ | ✓ |
| Data residency — NWA cloud | ✓ | Optional | — |
| Data residency — customer premises | — | Optional | ✓ |
| Local RAG (documents stay on-node) | — | — | ✓ |
| Outbound-only mTLS network | — | — | ✓ |
| Self-healing supervisor | Cloud-native | Cloud-native | 5-stage on-appliance |
| Customer support | Email (business hours) | Direct engineering | Direct engineering |
| Primary execution boundary | Managed cloud | Mixed | Private local |
| Pricing basis | System scope + usage | System scope + placement | System scope + infrastructure |
The system’s outcomes, capabilities, integrations, execution boundaries, evaluation requirements, usage, and operating responsibility. We define those together before you commit.
No. They are execution-placement profiles inside Cruuvo. Every system gets the same controls, evaluation, evidence, and improvement model.
AI usage is billed at Cruuvo’s published model-tier rate. The rate includes model inference and the infrastructure required to run it reliably—routing, networking, security, observability, retries, and capacity.
You choose one of four strategies: Open Weight ($), Smart Router — Open ($$), Smart Router — Private ($$$), or Frontier ($$$$). Smart Router automatically balances quality and price within the training-use boundary you select.
Every run records the model used, usage is visible in the Cruuvo workspace, and explicit run and account budgets stop work before silent overspend.
Yes. System definitions are portable and versioned. A reviewed release can change placement without turning Core, Forge, or Edge into different product objects.
Core runs in NW Agentic-managed cloud. Forge coordinates managed and private execution according to each capability’s declared boundary.
On Edge, data is stored on the on-prem appliance at your premises. Only retrieved chunks ever cross the network boundary; full documents stay on-node.
Yes. Audit logs and account data are exportable. Workflow definitions are portable across placements — a Forge workflow can move to Private Edge without rewriting the workflow.
Yes. Runs have explicit budgets, deadlines, and retry limits. The runtime checks authority and available budget before model calls and external effects.
Talk to us. For multi-system or multi-site operations, we structure pricing around shared capabilities, placement, and operating load.
We’ll map outcomes, capabilities, placement, evaluation, and run-rate so the commercial model matches the operation.