Technology
OpenQV: the authorization engine underneath the journey
OpenPaymentGPS shows customers where money is. OpenQV is the layer that decides whether an instruction submitted by an automated or intelligent system is permitted before value moves, and normalizes every bank and network event into one consistent language.
Before value moves
Twelve questions answered on every instruction
- 01
Who owns the automated system?
- 02
Who authorized the action?
- 03
What is it permitted to do?
- 04
Which account or payment method can it use?
- 05
What amount can it spend?
- 06
Which beneficiary can receive it?
- 07
Which payment rail should process it?
- 08
What compliance and risk controls apply?
- 09
Does the action require human approval?
- 10
What evidence must be preserved?
- 11
How is the transaction executed?
- 12
How is it reconciled?
Financial Action Gateway
A corporate treasury assistant, under policy
A $500 payment may proceed automatically. $5,000 asks for manager confirmation. $50,000 requires two authorized executives and enhanced beneficiary verification. A new or high-risk beneficiary pauses execution for compliance review.
Delegated authority
Grant for agent_2741
Treasury assistant acting for company_108. Adjust the instruction and watch the authorization decision change.
- auto_approve_below
- $1,000
- dual_exec_above
- $25,000
- ceiling
- $50,000
{
"agent_id": "agent_2741",
"principal_id": "company_108",
"action": "pay_invoice",
"beneficiary": "merchant_742",
"amount": 4800,
"currency": "USD",
"invoice_id": "INV-2026-817",
"cross_border": false
}{
"decision": "APPROVED_WITH_CONDITIONS",
"authorization_confidence": 99,
"maximum_authorized_amount": 5100,
"permitted_beneficiary": "merchant_742",
"human_approval_required": true,
"approvers_required": 1,
"recommended_rail": "RTP_INSTANT",
"evidence_record": "QV-347599",
"valid_until": "1970-01-04T00:00:00Z"
}Policy checks
- Agent identity attestedagent_2741 bound to principal company_108
- Action within delegated scopepay_invoice is a granted capability
- Currency permittedUSD allowed
- Amount within authority ceiling$4,800 of $50,000 ceiling
- Beneficiary verificationmerchant_742 on the approved beneficiary list
- Sanctions and AML screeningNo hits returned by the bank's existing AML system
- Geographic and corridor policyDomestic corridor within permitted region
Execution conditions
- Manager confirmation required before AI PayThrough executes
- Single-use credential scoped to beneficiary, amount and expiry
The agent never receives account or card credentials. AI PayThrough executes a single-use instruction bound to merchant_742, capped at $5,100, expiring 1970-01-04T00:00:00Z.
Operating baseline
We observe, normalize and present. We never become the rail.
OpenPaymentGPS correlates payment status from authorized sources. It does not become SWIFT, Fedwire, CHIPS or any other settlement rail, and inferred status is never presented as confirmed settlement.
Traceability
Material decisions, state changes and privileged actions can be reconstructed after the fact.
Least privilege
A person, service or device holds only the authority its role requires.
Fact before estimate
Confirmed events are separated from estimates, predictions and user assertions.
Fail safely
Uncertainty and unavailable sources produce explicit exception states, never fabricated success.
Additive corrections
No one alters a historical source event; corrections are appended and auditable.
Observable operations
Metrics, logs, alerts and case queues carry named ownership of every failure.
Who it is built for
How performance is measured
- Median exception-resolution time
- Connector availability and event-ingestion latency
- Trace completeness and reconciliation rate
- Enterprise user adoption and active cases
- Customer-service contacts per tracked payment
- Pilot conversion to recurring contract
Overlay, not replacement
Keep your payment infrastructure. Add a controlled gateway on top of it.
Authorization, policy, credentials and settlement stay under the bank's control.
Bank-grade requirements
- Private-cloud and on-premises deployment
- Bank-controlled encryption keys
- Regional data residency
- Zero-data-retention options
- No bank data used for model training
- Configurable human approval
- Role- and attribute-based permissions
- Tamper-evident audit records
- Model and provider governance
- Explainable authorization decisions
- Limits and beneficiary restrictions
- ISO 20022-compatible messages
- REST APIs, SDKs and webhooks
- Integration with existing fraud and AML
- PCI DSS-aligned credential handling
- SOC 2 and ISO 27001 readiness
- Operational resilience and failover
- Emergency revocation
