Within the evaluated policy
The supported checks find no reason to hold or reject. This does not guarantee a fraud-free payment.
AI PAYMENT SCREENING
Give agent purchases
clear boundaries.
Verify supported authorization artifacts, apply spending policies and return a recommendation with reasons before your integration executes a purchase.
From €90 / month minimum · Prices exclude VAT
REST API or MCP · Assisted integration
THE ENGAGEMENT
Know what goes in.
Know what comes back.
- INPUTS & COVERAGE
- Supported authorization artifacts, merchant restrictions, transaction and daily limits, and human-review thresholds. Protocol readiness and current deployment availability are separate.
- DELIVERABLES
- APPROVE, HOLD or REJECT, reasons and reviewable decision evidence. Your system controls payment execution. Human-review and governance features depend on the service package.
- GETTING STARTED
- Describe the agent, protocol, transaction volume and required behavior when a decision is held or the service is unavailable. Integration is quoted separately.
DECISION + REASONS + EVIDENCE
A control your team can explain.
Review before proceeding
The recommendation needs human attention or a control could not establish enough confidence.
Do not execute as requested
Authorization verification failed or the configured policy rejects the request.
The control must be enforced by your integration. An agent that can ignore the recommendation does not provide a reliable payment boundary.
PROTOCOL COVERAGE
Availability and readiness,
stated separately.
ACP, AP2 v0.2, x402, UCP and MPP have verification adapters. AP2 legacy is retained for compatibility. Supported modes and qualification differ.
ACPStripe Shared Payment Token · availability not checked
AP2 v0.2signed mandates · availability not checked
x402EVM exact scheme · availability not checked
UCPRFC 9421 signatures · availability not checked
MPPTempo credentials · availability not checked
AP2 legacydeprecated, superseded by AP2 v0.2 · availability not checked
Where this page can reach the service, each rail says what that deployment answers right now; where it cannot, it says so rather than assuming. Qualification is a separate question — see verification scope and qualification, or ask the service itself — that answer wins over this page.
An agent can call this directly. The same screening is published over the Model Context Protocol at api.trustedpai.com/mcp: five tools screen a payment, one per rail, and two cover what happens around a decision. Same engine, same keys, same limits. How to call it from an agent →
Deployment availability has not been checked in this page. Confirm enabled protocols for your account before integrating.
LET’S TALK ABOUT YOUR FLOW
Let’s scope your integration.
Describe the agent, protocol, transaction volume and required behavior when a decision is held or the service is unavailable. Integration is quoted separately.