Vitto Work Unit (VWU) Pricing Specification

Status: Draft Author: S202 Engineering Review Date: 2026-03-11 Type: Product Specification (design doc, not implementation)

1. What is a VWU?

A Vitto Work Unit is the outcome-based billing unit for Vitto's AI capabilities. Users pay for successful outcomes, not raw API calls or token consumption. This aligns revenue with user-perceived value and makes costs predictable.

Principle: A VWU is only charged when Vitto delivers a verifiable outcome. Failed tasks, retries, and internal overhead are absorbed by the platform.

2. Actions That Count as VWUs

Not charged:

• System prompt assembly, dossier hydration, memory extraction (internal overhead)
• Failed tool calls that are retried internally
• Rate-limited requests that return errors
• Health check / observatory endpoints
• SSE event delivery (infrastructure cost)

3. Pricing Tiers

Overage: Pro/Business users can purchase additional VWU packs:

• 100 VWUs: $8 ($0.08/VWU)
• 500 VWUs: $35 ($0.07/VWU)

Free tier purpose: Acquisition funnel. 50 VWUs = ~30 chat messages + 1 module generation. Enough to experience value, not enough to run a business.

4. Cost-per-VWU Target (Internal)

For viable margins, the platform cost per VWU must be below 50% of the charge price.

Current estimated costs (from Observatory data, S201):

• Simple chat: ~$0.003-0.01 LLM cost (well within margin)
• Module generation: ~$0.08-0.15 LLM cost (5 VWU = $0.25-0.29 revenue at Pro, marginal)
• Computer use: ~$0.05-0.10 per 5-min block (2 VWU = $0.10-0.12 revenue, tight)

Risk: Module generation and computer use are near break-even at Pro pricing. Mitigation: response caching (S201), scaffold fallback (reduces LLM calls), and model cost decreases over time.

5. Dispute Window

Window: 60 minutes from VWU charge.

If a user triggers an action and immediately undoes it (e.g., deletes a generated module within 60 minutes), the VWU is refunded. This prevents charging for accidental or experimental work.

Implementation notes:

• Each VWU charge has a chargedAt timestamp and status (pending | confirmed | refunded)
• After 60 minutes, status moves from pending to confirmed (cron job)
• Undo/delete actions within the window trigger automatic refund
• Refunded VWUs are added back to the user's monthly balance

6. Measurement Method

How Vitto detects "successful completion":

Counting rules:

1. One VWU per outcome, not per API call. If a chat needs 3 retries internally, it's still 1 VWU.
2. Compound tasks (e.g., "generate 5 modules and deploy") charge for each sub-outcome independently.
3. Background tasks charge when they complete, not when they start.
4. Cancelled tasks are not charged (no outcome delivered).

7. Data Model (Schema Sketch)

VwuCharge {
  userId: string
  spaceId: string
  action: string           // 'chat', 'module_generation', 'deployment', etc.
  weight: number           // VWU weight for this action
  status: 'pending' | 'confirmed' | 'refunded'
  chargedAt: Date
  confirmedAt?: Date       // Set after dispute window closes
  refundedAt?: Date
  taskId?: string          // Link to the task that generated this charge
  metadata: {
    actualCostUsd: number  // Internal LLM cost for margin tracking
    model: string
    tokensUsed: number
  }
}

VwuBalance {
  userId: string
  tier: 'free' | 'pro' | 'business' | 'enterprise'
  monthlyAllowance: number
  used: number
  overage: number
  billingCycleStart: Date
  billingCycleEnd: Date
}

8. Open Questions for Product Review

1. Should free-tier users see VWU counts? Showing "3/50 VWUs used" creates awareness and upgrade pressure, but might feel transactional.

2. Should module generation charge per-module or per-generation? Per-module (current proposal) is fairer but creates billing complexity for partial failures (3/5 modules succeed).

3. Should computer use be time-based or action-based? Time-based (current: per 5-min block) is simpler but penalizes slow connections. Action-based is fairer but harder to measure.

4. Should proactive alerts count as VWUs at all? They're initiated by Vitto, not the user. Charging for them might create negative perception ("Vitto is spending my credits without asking").

5. Enterprise custom pricing: Should enterprise include unlimited VWUs, or always be metered with volume discounts?

This is a product specification. Implementation requires schema creation, billing integration, and frontend VWU counter. Estimated effort: 2-3 sessions after product approval.