Resolve per-call pricing: discovery is free; execution is pay-per-call. Agents should estimate route cost before execution, choose the correct rail, keep cost separate from AN Score quality, and inspect spend plus selected provider after the call.

Pricing authority

Per-call pricing only works if the route is explainable.

Discovery is free. Execution is per-call. Before an agent spends money, Resolve should explain the provider path, credential rail, estimate, and constraints that shaped the route.

Canonical pricing Credential paths

Provider cost

The upstream cost of the selected provider or route.

Credential mode

Rhumb-managed, BYOK, Agent Vault, wallet-prefund, and x402 rails have different economics and control boundaries.

Route decision

The cheapest provider is not always the best route; cost is disclosed as a constraint, not hidden inside quality.

Operator policy

Budgets, limits, and allowlists can constrain route choice before execution.

Call sequence

Estimate first. Spend second.

The agent should not discover price by surprise after execution. The estimate is part of the route contract.

  1. 01Discover and resolve the capability.
  2. 02Estimate route, readiness, and expected call cost.
  3. 03Choose governed API key, wallet-prefund, x402, BYOK, or Agent Vault.
  4. 04Execute only after the route and rail are explicit.
  5. 05Inspect spend, selected provider, and alternatives considered after the call.

Neutrality rule

Cost can change the route. Cost must not rewrite the score.

AN Score remains a quality prior. Cost is an execution constraint and should be disclosed in the route explanation, not smuggled into service quality.